Zurich Patch 12

The Zurich Patch 12 release contains important problem fixes.

Zurich Patch 12 was released on August 07, 2026.

• Build date: 07-31-2026_1556
• Build tag: glide-zurich-07-01-2025__patch12-07-16-2026

Important:

For more information about how to upgrade an instance, see ServiceNow upgrades.

For more information about the release cycle, see the ServiceNow Release Cycle.

Note:

This ServiceNow AI Platform® major family release is now available in ServiceNow's Regulated Market environments. For more information about services available in isolated environments, see KB0743854.

For a downloadable, sortable version of the fixed problems in this release, click here.

Overview

Zurich Patch 12 includes 396 problem fixes in various categories. The chart below shows the top 10 problem categories included in this patch.

Figure 1. Top 10 problem categories

Security-related fixes

Zurich Patch 12 includes fixes for security-related problems that affected certain ServiceNow® applications and the ServiceNow AI Platform®. We recommend that customers upgrade to this release for the most secure and up-to-date features. For more details on security problems fixed in Zurich Patch 12, refer to KB3140941.

Changes in Zurich Patch 12

Configure

Configure Dynamic Guidance to be available in the Help Center and Now Assist panel. When sn_dyn_guidance_user role is assigned, it also includes the genai_admin role.

Note:

The genai_admin role does not grant administrative privileges.

Using Dynamic Guidance

Learn how to invoke Dynamic Guidance and use it effectively to enhance your experience. Starting with Dynamic Guidance version 28.4.3, the genai_admin role is automatically included when the sn_dyn_guidance_user role is assigned. The genai_admin role does not grant administrative privileges.

Notable fixes

The following problems and their fixes are ordered by potential impact to customers, starting with the most significant fixes.

Problem	Short description	Description	Steps to reproduce
AI Search for Service Portal PRB2038055 KB3139447	The Now Assist Self Service widget breaks the full page experience search in portal	After upgrading to Zurich, when the user searches from the Employee Center portal, the full page experience fails and no results are displayed. Instead, the widget throws a server-side JavaScript error: '$sp is not defined'.	Refer to the listed KB article for details.
Key Management Framework (KMF) for Platform Encryption PRB2058369 KB3140571	Midserver is unable to fetch credentials after upgrading to Zurich or Australia	In certain versions, there's a Unified Secrets Gateway (USG) service for credential management. During the upgrade to those versions, a system trigger script is designed to automatically execute and populate the sys_​secret_​identity _​group_​member table with the MID Server identity group mappings required for USG authentication. However, this trigger fails to complete successfully, leaving the table incompletely populated. As a result, the MID Server can't authenticate with USG and fails to retrieve credentials.	Refer to the listed KB article for details.
Now User Experience PRB2038702 KB3134148	A scoped public UI page isn't accessible without a login unless 'Name' is used vs a scoped endpoint in a sys_public record	It should be accessible without a log in, so a user can complete the authentication from Outlook.	Refer to the listed KB article for details.
System Events PRB1969068	The 'Events process 0' job yields to memory pressure, causing event processing delays	When node memory pressure is high (live set ≥ ~91%), the JobYieldCheck mechanism triggers a yield on events process 0 despite it being a priority 25 (high priority) job. High-priority jobs should not be subject to automatic yield throttling, as this directly delays critical event processing and can result in P1 incidents. The memory pressure itself may be transient or difficult to diagnose quickly — heap dumps often show no obvious culprit, and identifying the root cause of elevated memory usage takes time. During that window, events process 0 is repeatedly yielded, stalling event processing pipelines that users depend on for time-sensitive operations. JobYieldCheck WARNING Job=events process 0 yields due to memory pressure logged on affected nodes. Node memory is sustained at 91–93% of max, with live set at ~91.44% (~1.28GB). Glide.memory.watcher logs no active transaction warnings alongside persistent memory pressure status = true. There's only a single job visible in the queue during the yield window, yet throttling still triggers. However, users expect priority 25 jobs (events process 0) should not be yielded under memory pressure conditions. Yield throttling should be restricted to lower-priority workloads.
Table Rotation PRB1959672 KB3071677	Shard tables aren't created on an initial plugin install	When a plugin is manually installed, shard tables aren't created automatically.	Refer to the listed KB article for details.
UX Framework PRB2027095 KB3102030	The latest component asset isn't selected when multiple asset associations exist in sys_​ux_​lib_​ component_​m2m_​asset,​ which impacts AI summary cards on UI Builder workspaces	When the AI summary card component was upgraded, its asset name changed from now-ai-summary-card/index to uxc-generative-ai/index. The upgrade doesn't clean up the old asset association, so the sys_ux_lib_ component_m2m_asset table ends up with two associations for the same component (old and new asset). The component-to-asset selection query in Glide​Ux​Component​Def​Provider .​get​Asset​Names​By​Component​Sys​Ids()​ had no ordering and relied on default DB ordering. When duplicate associations exist, this could return the stale now-ai-summary-card/index asset, which is incompatible with the latest Platform AI Agents and Skills app. As a result the AI summary card fails to load on UI Builder workspaces. Classic UI uses a different rendering path, so it is unaffected.	Refer to the listed KB article for details.

All other fixes

Problem	Short description	Description	Steps to reproduce
Activity Stream PRB1991852	Base64 coded attachments can't be loaded in the activity stream in the workspace	This likely has to do with processing large content.	Navigate to any record page. Create an email with a base64 encoded image. Set the email to 'sent' status. Navigate to the record and select Show more for the email. Expected behavior: The full body of the email is displayed. Actual behavior: The email is loading for a long time and eventually crashes the page.
Activity Stream PRB2040417	The Copy button on the VTB view can't access copyJournalContent or GlideUIDefault	The content doesn't get copied to the clipboard API and doesn't show the 'Copied to clipboard' notification.	On any instance, navigate to All > Visual Task Boards. Create a Freeform Board. Add a card on the newly-created board. Open the card and add a work note or comment. Select the Copy journal content button. Expected behavior: The content is copied to the clipboard API and shows a 'Copied to clipboard' toast. Actual behavior: The content doesn't get copied to the clipboard API and doesn't show the 'Copied to clipboard' notification.
Agent Chat PRB2021594	An agent chat box isn't enabled when work is offer re-routed post blind queue transfer	In most recent Australia release, the CSM Workspace agent isn't able to type messages in the chat input box after they are re-routed an interaction that was rejected by a previous agent. The chat box should be enabled after re-routing acceptance.
Agent Chat PRB2034807	UXF changes for inbox DOM	Assist UXF with the changes being implemented to remove inbox from the DOM when navigating away from workspace.	Open ServiceNow Workspace with Agent Inbox loaded. Navigate away from the workspace to another page/module. Inspect the DOM. Observe that the inbox component remains in the DOM instead of being removed on navigation.
Agent Chat PRB2051353	Agent presence indicator should be shown in the 'Transfer to Agent' list for third party	When searchTargetList is set, a list of agents is shown, but there's no presence indicator.	Create and offer a third-party messaging interaction to a CSM Agent. Select the Transfer to Agent quick action. Expected behavior: It shows a list of agents along with presence indicator when searchTargetList is set. Actual behavior: It shows a list of agents with no presence indicator when searchTargetList is set.
Agent Chat PRB2052028	Integration users can't read sys_cs_conversation_member for creating wrap-up segment	The wrap-up segment creation fails with a 400 status and the error message: 'Failed to create wrap-up segment. Segment not found in implementations for extension point: interactionSegment'.	Create a third-party messaging interaction. Offer to a CSM Agent. Call REST API to create a wrap-up segment for that interaction and agent. Expected behavior: The wrap-up segment is created and the API returns success. Actual behavior: The wrap-up segment creation fails with a 400 status and the error message: 'Failed to create wrap-up segment. Segment not found in implementations for extension point: interactionSegment'.
Agent Chat PRB2055561	Rename Now Assist, Moveworks and AI experience to ServiceNow Otto	The Sparkle button for CRR should be replaced with the new Otto button. CRR button's tooltip 'Write with Now Assist' should be replaced with 'Write with AI'. Finally, the Sparkle icon and its label in the recommendation modal should be replaced with the Otto icon.
Agent Chat PRB2058031	Add a missing flag on 'Requires ACL authorization'	A fix has ensured that the 'Requires ACL authorization' flag on 'Conversation Server' graphQL schema is set to true. However, due to a discrepancy in syncing branches, only the ACL changes added are live and not the graphQL schema changes.
Agent Chat PRB2063949	Some messages are hidden on the agent's side when hideControl is true and contains searchText
AI Agents (Glide Family) PRB2034699	GenAIMetadataM2MDaoImpl createAIAGenAIMetadataM2M doesn't assert that the genAILogId parameter is not NULL	This causes a full table scan of the sys_gen_ai_log_metadata table. This code is invoked by the AsyncMessageProcessor.
AI Agents (Glide Family) PRB2037758	AI Agents randomly throw a tool execution error: 'Failed to execute async FDIH tool'	At any random tool call, the AI Agent throws: 'Sorry, there was a problem on my side trying to complete this request. Try asking again later'. In logs, it shows 'FDIH async tool call exception'.	Log in to an instance. Invoke the Quote AI agent. Input a multi-paragraph, conversational prompt. Observe that, at any random tool call, it throws: 'Sorry, there was a problem on my side trying to complete this request. Try asking again later'. In logs, it shows 'FDIH async tool call exception'.
AI Agents (Glide Family) PRB2051367	TTL isn't honored in sys_​og_​conversational _​cache_​configuration in the absence of any other invalidation rule	The persona section doesn't show the new memories immediately, but they're visible after 24 hours.	Set up LTM extraction on Glide (enable sys_prop use_new_ltm). Set up DARE with memory enabled (in pipeline.yaml, set enable_user_persona). Create a conversation with some details that are a candidate for LTM extraction (for example, 'I like cycling'). Verify that memories are extracted (the ltm job runs every hour). After new memories have been extracted, check the persona section. Observe that it doesn't show the new memories. However, they will be visible after 24 hours.
AI Agents (Glide Family) PRB2059351	Record scope should override the user session scope	For AI Agents, when any record of a user instance is used, the record scope should override the user session scope of the MSP as per the domain separation policy. However, that's not happening. For example, when an AI Agent with a retriever tool is triggered by any record in a user instance (like C1), it fetches KBs/results from all the user instances under that MSP.
AI Agents (Glide Family) PRB2059390	Role masking isn't unset when running a workflow
AI Experience Framework - Glide PRB2061584	True-up the AI UX Builder Store app
AI Gateway - Security PRB2052119	As part of the name change, update error strings in MCPOAuthConstants.java that reference 'Now Assist' to 'ServiceNow Otto'	The changes will be made in Zurich and Australia. Also, an additional change is needed for the description in glide/glide-ai-security-scan.
AI Search (Glide) PRB2052168	No_answer Genius Result is streamed intermittently	A Genius Result rarely shows, but it should not show at all.	Navigate to /sp. Search for something that gives a no_answer result, like 'baseball world series'. Observe that a Genius Result rarely shows. It should not show at all.
AI Search (Glide) PRB2054276	The fix script shouldn't backfill non-default sources	When the user upgrades to an instance with additional sources in the multi-content feature, the additional sources are marked as 'Included'.	Upgrade an instance without additional sources in the multi-content feature to a version with the feature. Observe that the additional sources are marked as 'Included'.
AI Search (Glide) PRB2054932	In Moveworks API V2, support linking ServiceNow Security Center (SSC) to a search profile	This is a product update.
AI Search (Glide) PRB2054938	For the Moveworks API V2, return chunks when passing an additional flag	This is a product update.
AI Search (Glide) PRB2058047	Indexing with multiple semantic indexing configurations with the same name on different models isn't working	When multiple semantic index configurations are created on the same datasource with the same semantic field name but different embedding models, only the first configuration is loaded into the active semantic index field cache. Subsequent configurations are silently ignored, so ingestion and search only use one embedding model for that field. Only one embedding model is used for indexing and search when multiple semantic index configurations share the same semantic field name. Additional configurations with the same semantic field name aren't visible in get​Semantic​Index​Field​Mapping()​.​ No error or warning is logged when the duplicate-name configurations are skipped. Thus, multi-embedding model support for a single semantic field is broken. Configurations are silently lost during cache population. This affects ingestion, search, and any callers that rely on get​Semantic​Index​Field​Mapping()​.​	Create two active ais_​semantic_​index_​configuration records on the same datasource with the same semantic_field_name but different embedding_models values. Add valid component fields via ais_semantic_component_field for each record. Set a valid semantic_​snippetization_​configuration.​ Flush the datasource object cache (Ais​Configuration​Cache​Manager .​flush​Datasource​Object​Cache()​)​.​ Call Ais​Configuration.​get()​.​ get​Semantic​Index​Field​Mapping ('kb_​knowledge',​'kb_​knowledge')​.​ Observe that only one SemanticFieldConfiguration is returned for semantic_search and the other is silently dropped.
AI Search PRB1875924 KB3137872	AI search ingestion sticks when the 'Size in Bytes' column has an empty value in sys_attachment for an attachment that has a size more than 25MB		Refer to the listed KB article for details.
AI Search PRB1902419	The conversation ID isn't logged to the sys_search_event table		On an AI Search-enabled instance, start a new chat with Virtual Agent (VA). Enter 'what is spam'. Flush queued signals. Open the sys_search_event table, and if necessary, add the Conversation ID field to the list/form. Expected behavior: Since it was from VA, there should be a conversation. Actual behavior: The conversation ID is empty.
AI Search UX PRB2004411	Refactor the synthesized Genius Result presence filter from Data Broker into the Java layer	Currently, the logic to conditionally execute the synthesized Genius Result — specifically, the filter that checks whether a Genius Result is present before execution — lives in the Data Broker transform layer (UX layer). This means the filter isn't automatically applied across all clients and must be manually maintained per-client in the Data Broker. This logic must be moved into the Java layer so that it is applied universally and automatically for all users, reducing duplication and the risk of inconsistent behavior across different consumers of the Search API.	Review the Data Broker transform at sys_ux_data_ broker_transform_ 0c735fb4530422104 accddeeff7b12dd.xml#L44. Check the conditional filter that gates execution of the synthesized Genius Result on its presence. . Notice that this filter only applies to clients using this specific Data Broker transform; it isn't enforced at the Java/API layer. A user that bypasses or doesn't use this Data Broker transform doesn't have the filter applied, resulting in potentially inconsistent or unintended behavior around synthesized Genius Result execution.
AI Search UX PRB2014714	Ignore Genius Result's limit for citations in sync mode		On an AI Search enabled instance with synthesized response configured for portal, navigate to a portal. Complete a query that yields a synthesized response. Set the system property 'glide.​ais.​query.​disable_​async_​mode' to true. Do same search in portal. The user should still get a synthesized response. Open the search application (presumably NAVA) and set the Genius Result limit to one. Do the same search in portal. Expected behavior: Users still get a synthesized response. Actual behavior: No Genius Result renders because of no citations.
AI Search UX PRB2038703	Instance creation failure for com.​glide.​search.​ graphql.​query.​Suggestions from service portal's typeahead AIS Suggestions API	When something is searched for in the portal's typeahead, no suggestions appear and AIS Suggestions API calls in the network tab return 'Instance creation failure for: com.​glide.​search.​ graphql.​query.​Suggestions',​ and 'DataFetchException'.
AI Search UX PRB2054213	URLs with special characters like $ break streaming for non-VA clients	Streaming starts successfully, but it stops partway through and the full answer is never rendered.	Open an AIS-enabled instance. Make sure 'Synthesized Response' is enabled in portal. Navigate to the portal with Now Assist enabled. Search for 'How can I reset my password?'. Expected behavior: The full answer streams successfully. Actual behavior: Streaming starts successfully, but it stops partway through and the full answer is never rendered.
AI Search UX PRB2054952	Rename Now Assist to Otto within AI Search components	This is a product update.
AI Search UX PRB2056856	Replace the Sparkle icon animation with the Otto animation for synthesized genius results (GR)		Open an instance with multi-content GR configured. Navigate to portal/global search. Search for a query that returns a synthesized GR. Check the loader in the GR component when the loading state is showing. Expected behavior: The loader state shows the Otto animation. Actual behavior: The loader state shows the Sparkle icon animation.
AI Search UX PRB2057349	Rename Now Assist to Otto for typeahead suggestions	This is a product update.
Analytics Data API PRB2039649	The Platform Analytics dashboard ignores the data source filter	The 'Is one of' operator doesn't function as expected on the source. The issue is reproducible in Australia with the Data Visualization plugin version 29.1.0 and in Zurich with version 28.5.0. However, the functionality works as expected in Yokohama with version 28.0.29 and in Zurich with version 28.4.5.	Open an instance. Create an indicator scorecard visualization. Select Number of Incidents as the data source. Apply a source condition using the 'Is one of' operator. Select the required priority values. Expected behavior: Only the selected priority is reflected. Actual behavior: All priorities are reflected in the indicator scorecard visualization.
Analytics Export API PRB2021342	The email layout in an Platform Analytics scheduled export doesn't match the received email format	Platform Analytics scheduled reports aren't honoring the email body configured when the field omit_if_no_records is true.
Application Install Engine PRB2056922	Install engine reads the dependencies from package.json with a fall back to the previous way when not available	This is a product update.
Application Manager PRB2022268	The application manager sys_app_version displays duplicate records for the same application and version, which is causing the app to be 'Installation blocked'	As part of the AI testing, it's been observed that the app installations are blocked. The sys_app_version displays duplicate records for the same application, which is causing the app installations to be blocked, despite the app versions being available as well as the license checks having successfully completed.	Navigate to an instance. Navigate to App Manager. Open any app (for example, sn_ai_itsm_cont) which is licensed and validated on CI (usageanalytics) prod, but has yet showed 'Not Licensed'. Select Install. It just fails with 'Installation blocked' on that app itself. Open sys_app_version table. Check the app/scope id. Expected behavior: It should have only one record for that app and a specific app version. Actual behavior: It has multiple records for the same app and app version.
Application Manager PRB2057729 KB3140027	Users can't publish custom-scoped applications after upgrading to Zurich or Australia	After upgrading to Zurich or Australia, there's an issue that prevents new application versions from being published for custom scoped apps in ServiceNow Studio. Selecting the Publish button for any custom application results in a publish failure with one of the following errors: 'Vendor information was not found, upload function is disabled for this instance' or 'Did not receive publishing information successfully. Cannot read properties of undefined (reading 'toString')'.	Refer to the listed KB article for details.
Application Rationalization PRB1977007	Generate Assessment takes too long to load because of excessive sys_user_group records	The issue occurs when there are a lot of records in the sys_user_group table. The sn_apm.Assessment AjaxProcessor script include calls the getInitData method, which contains some code that queries all sys_user_group records. There is no limit, so instances with many sys_user_group records may encounter the following issues: first, excessive memory usage by the userGroupRecords object, causing node performance issues and even a node restart. Second, excessive payload returned by the function to the application node and eventually to the client browser, causing slowness.
Asset Management Common PRB1982019	Service locations on stockroom aren't added through the MRA pop-up	The success message is shown, but records aren't added under the related list.	Log in to an instance. Navigate to the stockroom record in workspace. Select the Service Locations related list. Select the Add button. Select two locations in the MRA pop-up. Select Add. Expected behavior: Locations are added under Service Locations. Actual behavior: Locations aren't added under Service Locations.
Authentication Factors PRB2037251	Update interactions post-identification and authentication	The 'Interactions' table has only 'Guest' resolved. It should have the user reference resolved post-identification and authentication.
Authentication PRB2037225	Third-party access token authentication fails	This is a product update.
Automated Test Framework (ATF) PRB1942035	sn_cicd API for test run doesn't start the test in cloud runner when run_in_cloud is set to true	The test remains in the waiting state until a scheduled client runner is opened. It fails after waiting for ten minutes due to timeout. No record is found in sn_atf_tg_sn_boq for the test run submitted.	Open an instance with ATF Cloud runner enabled. Log in as admin. Navigate to REST API Explorer. Navigate to namespace: sn_cicd and API name: CICD ATF Test Execution API. Select Start a ATF, given a test ID to open the API to run a test Enter the test_sys_id of any UI based test. Enter run_in_cloud = true. Leave is_performance_run as blank. Submit the request. Navigate to the test page for the sysID. Check the 'Test Results' tab. Open Browser Orchestration Queue (sn_atf_tg_sn_boq). Look for the record created for the test run. Expected behavior: The test run should be queued in sn_atf_tg_sn_boq. The test should run in cloud and complete. Actual behavior: The test result shows 'Waiting'. The test remains in the waiting state until a scheduled client runner is opened. It fails after waiting for ten minutes due to timeout. It's also not queued in Browser Orchestration Queue. No record is found in sn_atf_tg_sn_boq for the test run submitted.
Automated Test Framework (ATF) PRB2017949	There's stuck execution trackers due to Automated Test Framework (ATF)'s new feature metadata tracing	For invalid metadata table names in sys_traced_metadata records, the generateMissingPayloadHashes throws an error and causes the final payload hash calculation execution trackers to be stuck.	Ensure that metadata tracing is on. Run a test on a table with a long name (such as 'sys_og_conversational _cache_configuration', more than 40 characters). Observe that the sys_traced_metadata records truncate the long table name (sys_traced_metadata records save table names up to 40 characters whereas metadata table names can be up to 80 characters in length). Observe at the end of the test, there's stuck sys_execution_trackers while generating payload hashes for the sys_traced_metadata records because the truncated metadata table name is invalid. It thus throws an error.
Canonicalization Data Services (CDS) PRB2053470	Data uploaded to cds_server_staging table is blocked within instances	This issue occurs in instances where the property 'glide.cmdb.canonical.URL' is set to https:​/​/​<instance_​name>​.​service-​now.​com/​ and ends with '/'. Unfortunately, that is the base instance value, and it affects the upload portion of CDS.
Change Management PRB2035924 KB3093597	Standard changes created from a problem aren't linked to the parent task	The issue is reproducible if users clear the 'Two step' checkbox under Change > Administration > Standard Change Properties and then perform cache.do > Clear Cache. A standard change created from a task (incident/problem) aren't linked to a parent task. If users check conflicts before saving a child change from a project requested item, it won't attach the change to the parent item.
Change Management PRB2053039	In the 'Change risk assessment' answer generator skill flow, replace the Now Assist terminology with Otto rebranding	The suggestion message should contain 'Check answers generated by AI for accuracy'.	Navigate to SOW > Change Request. Select Assess Risk in the 'Overview' tab. Select Generate Answers. Wait until risk assessment completes answer generation. Open risk assessment. Observe that the suggestion message doesn't contain 'Check answers generated by AI for accuracy'.
Client Scripts PRB1956934	g_form.getActionName returns 'none' when adding a file in a File Attachment field		Create an OnSubmit client script for incident table. Add a File Attachment field in the incident form. Select Click to add... in the File Attachment field. Observe that g_form.getActionName returns 'none'.
Condition Builder PRB2054935	Create a sys property to determine the Glide version for condition builder	This is a product update.
Configuration Management Database (CMDB) PRB1994587	A race condition between IRE LazyWriter and Deletion Strategy causes the configuration item (CI) state to flip incorrectly	A CI's state is flipping from installed to stale during Discovery (delete strategy execution). A race condition occurred between IRE's asynchronous batch update of the last_discovered field using 'db.executeLazy()' and the execution of the Deletion Strategy. The LazyWriter delay caused the Deletion Strategy to process stale 'last_discovered' values, marking valid CIs as retired despite their presence in the payload.
Configuration Management Database (CMDB) PRB2017559	The health configuration cache is ineffective due to a shared cache eviction	The getHealthConfiguration() method in HealthConfigManager.java is not getting effective cache hits. The current implementation caches per (metricSysId + className + domainId) key, creating a few hundred cache entries. Since this is a shared cache, entries get evicted before they can be reused, resulting in repeated DB queries for records that never change during processing.	Create health configuration rules and observe the cache entry through cache_inspect.do. Create a pressure on the system to cause cache eviction. Observe in the code path that the cache is rebuilt every time there is an eviction. The caching path is controlled via sys_property and only applies to the non-domain path.
Configuration Management Database (CMDB) PRB2020295	INSERT_NOT_ALLOWED_FOR_SOURCE incorrectly classified as a partial error causes the cmdb_ire_partial_payloads table to bloat and an out of memory (OOM) error	When an IRE data source rule blocks inserts for a class and data source, affected records are incorrectly saved as partial payloads. Since the block is permanent, these partial payloads can never complete — each retry creates a new entry, causing the cmdb_ire_partial_payloads table to grow unboundedly. If the partial payload contains a large number of related or lookup items, this can also lead to OOM on app nodes.
Configuration Management Database (CMDB) PRB2021367	Implement totals calculation for the group view when applying a class filter qualifier configuration
Configuration Management Database (CMDB) PRB2030304	OrphanProcessor skips CIs and reprocesses records after batch timeouts due to unordered HashMap iteration	When the CMDB Health orphan metric job times out MID-run on large datasets, it saves an incorrect resume position and silently skips CIs on the next run. As a result, orphan CIs appear healthy in the dashboard even though they are actual orphans, leading to inaccurate health scores.	Set up a class with a large enough population of orphan CIs such that processing exceeds five minutes. Trigger the CMDB Health orphan batch job. Allow the job to run until the five-minute batch timeout fires MID-loop. Observe the checkpoint value stored in MetricProcessorStatus for that class — it's an arbitrary sys_id, not the lexicographically highest one processed. On the next scheduled run, observe that orphan CIs with sys_ids smaller than the stored checkpoint aren't evaluated and not reported as failures.
Configuration Management Database (CMDB) PRB2052583	References to 'Now Assist' for the duplicate CI remediator should be renamed to Otto
Configuration Management Database (CMDB) PRB2055214	AI should be used to suggest options for de-duplication task/template remediation	This is a product enhancement.
Configuration Management Database (CMDB) PRB2055216	Related items handling for de-duplication process should be improved	This is a product enhancement.
Contract Management PRB2040201 KB3140742	Assets covered on a contract aren't added through an Multiple Record Associator (MRA) pop-up	In the 'add action' over MRA pop-up, there's a slight change by the platform. It is now performing a GlideRecordSecure insert instead of GlideRecord. It is validating for create and write ACLs on clm_m2m_contract_asset.asset and clm_​m2m_​contract _​asset.​contract fields. There's ACLs where it allows the respective operations only if they are blank. This is clearing out the Contract and Asset field values during creation. So, records which are created don't have these fields and aren't displaying in the related list.
Customer Service Management PRB2002976 KB3141009	An advanced qualifier reference that uses 'current' isn't functioning as expected in CSM Workspace	Advanced reference qualifiers that rely on the 'current' object fail when templates are applied in Workspace. However, the same functionality works as expected in 'Classic UI (UI16)'. During template application in Workspace, the platform passes corrupted or incomplete serialized data to 'GlideRecordEncoder.decode()'. As a result, the current GlideRecord is returned without field values.	Refer to the listed KB article for details.
Database Persistence - Data Management PRB1884666	TableCleaner. deleteAttachments() has an infinite loop and is unable to delete encrypted records		Insert an encrypted attribute into sys_attachment. Invoke TableCleaner. deleteAttachments() via a script.
Database Persistence - Data Scale PRB1942731	If sequences exist to reduce redundant queries, SequenceAtomicCounter.get could cache
Database Persistence - Graph PRB2026586	toLower() and toUpper() functions aren't supported in WHERE and RETURN	After creating a query with GraphQueryBuilder, users of GraphQueryBuilder API should be able to get the encoded query representation as a transferable serialization of the query. Cypher's toLower() and toUpper() functions aren't supported anywhere in the builder — neither in WHERE predicates nor in RETURN projections.
Database Persistence - Graph PRB2026589	COUNT(DISTINCT x) can't be represented with GraphQueryBuilder API	After creating a query with GraphQueryBuilder, users of GraphQueryBuilder API should be able to get the encoded query representation as a transferable serialization of the query. Aggregate('COUNT', x) only ever emits COUNT(x). There's no way to express COUNT(DISTINCT x), which is the only correct answer when multiple paths can reach the same node.
Database Persistence - Graph PRB2026590	ORDER BY isn't stored in Encoded QueryModel	After creating a query with GraphQueryBuilder, users of GraphQueryBuilder API should be able to get the encoded query representation as a transferable serialization of the query. RETURN alias — the AS clause is dropped when using buildEncodedQuery(), so any consumer of the encoded query has no way to know what column name the user asked for. ORDER BY / LIMIT / SKIP — these have no representation in QueryModel at all, so they're dropped whole on round-trip.
Database Persistence - Graph PRB2040734	Sys_class_name doesn't come as property from getForTables
Database Persistence - Graph PRB2051205	Inline node properties in the cypher are dropped in the encoded query when using with useCypher()	If there are inline properties in the cypher passed to useCypher, the returned encoded query doesn't contain that query.
Database Persistence - Graph PRB2052676	Cache flushes are triggered, even when the cache is not used/populated	The number of cache flush messages for the table 'cmdb_rel_ci' is in the 80 million+ range every hour. In a way, it's making 80 million inserts into the sys_cache_flush table. 'cmdb_rel_ci' is paired with 'graph_cmdb_rel_type_cache'. The cache graph_cmdb_rel_type_cache isn't enabled, but the cache pairing is still in a static block and is active.
Database Persistence - Graph PRB2053774	Support initializing GraphQueryBuilder with a cypher query string via the JavaScript API	This is a product update.
Database Views PRB1881428 KB2062907	A MariaDB database view using INSTR could result in a syntax error when the function is rewritten and executed on RaptorDB	This happens because of a related > 0 or != 0 comparison. When executed on MariaDB, INSTR() returns an integer and the comparison is to an integer. However, on RaptorDB, the rewrite results in a function which returns a boolean. Due to RaptorDB's strong typing semantics, a boolean can't be compared to an integer, resulting in the syntax error.	Refer to the listed KB article for details.
Data Privacy (Classic) PRB2031271	Rollback of a scheduled anonymization job doesn't refresh the activity stream UI for Journal fields (work_notes, comments)	When a scheduled Data Privacy anonymization job is rolled back, the Journal field values (work_notes and comments) are correctly restored in the sys_journal_field table. However, the activity stream UI continues to display the anonymized values because sys_activity table (used by Service Operations Workspace) and sys_​history_​set/​sys_​history_​line (used by UI16 List and Form views) aren't refreshed upon rollback. The Description field doesn't exhibit this issue. This is inconsistent behavior introduced when journal type anonymization support was added.	Open an instance with Data Privacy enabled. Configure a scheduled anonymization job targeting Journal fields (work_notes, comments) on the Incident table. Run the anonymization job. Confirm the fields are masked in both the database and the UI. Execute a rollback of the anonymization job. Verify in sys_journal_field that the original values have been restored. Navigate to the affected Incident record in both UI16 and Service Operations Workspace. Observe that the activity stream still displays the anonymized values even though the database was reverted.
Data Privacy (Classic) PRB2039332	Increase the maximum size limit of the incoming request to Data Privacy APIs from 1M to 10M	The Data Privacy API only supports up to 1M character long prompts. However, a simple prompt on Virtual Agent may result in prompts larger than 1M characters, which causes the prompt's response to get stuck. Prompts of extra large size should be processed by the Data Privacy API.
Data Privacy (Classic) PRB2040595	Post-clone anonymization of child workers is incorrectly cancelled due to parent job type mismatch	The post-clone data privacy anonymization job uses a parent-child architecture; one parent job distributes work to multiple child worker jobs that process tables in parallel. The issue is in how the child workers look up their parent job record. In the post-clone flow, the parent job is stored in a different table type than the child workers' lookup code expects. Because the lookup fails with 'data protection job is not valid,' each child worker is cancelled immediately.
Date Picker PRB2029576 KB3061444	There's an issue with GlideDateTime.setDisplayValue	Any script that calls setDisplayValue() with an ISO-format string (yyyy-MM-dd HH:mm:ss) on an instance where glide.sys.date_format is set to anything not in FORMAT_LIST (for example, dd-MM-yyyy, MM/dd/yyyy, etc.) produces incorrect dates in Zurich.	Refer to the listed KB article for details.
Decision Graph for Context Engine (Family) PRB2056900	Glide changes for machine learning integration needed for Distinguished Encoding Rules (DER) mining	This is a product update.
Decision Graph for Context Engine (Family) PRB2056902	Instrument 'Ask For' approval action Instance Observer (IO) during flow execution for context graph tracing	This is a product update.
Decision Graph for Context Engine (Family) PRB2056905	Rate-limited flow trace eligibility evaluation	This is a product update.
Decision Graph for Context Engine (Family) PRB2056908	Scaffolding for the decision-graph module	This is a product update.
Decision Graph for Context Engine (Family) PRB2056910	Cache decision weightage for a low-latency runtime lookup	This is a product update.
Decision Graph for Context Engine (Family) PRB2056912	Adding or deleting a feature dictionary should make mined patterns inactive	This is a product update.
Declarative Actions PRB2055766	The related list UI actions to update rows are failing on m2m if the field being updated is read-only	The UI action calls AssetAutomationAPI. completeAssetDisposalTask. This has current.update, which fails. If the read-only flag is removed from the Stage field in the sn_hamp_m2m_ hw _asset_disposal table, then it works.	Navigate to Hardware Asset Workspace > Inventory > Disposal order. Create a disposal order. Navigate to the related list. Add an asset to the disposal order. Open Verify task from the related list. Select the row. Select Verify. Observe that the action fails to update.
Developer Sandboxes PRB2054240	The scheduler claim mutex (sys_mutex) isn't sandbox-aware and forces DSB nodes to contend for a cluster-wide lock, causing scheduled-job pickup delay		On a multi-node instance (50+ nodes), turn on Developer Sandboxes. Create at least 20-30 sandboxes. Make sure multiple isolated sys_triggers exist in the sandboxes. Pull stats.​do?​include=​otel.​scheduler* on the sandbox's node. Observe that claim_lock_time averages above one second (expected ~13ms), jobs_lateness averaging 300+ seconds, and worker capacity used is very low. Compare against a controller node on the same instance. Observe that claim_lock_time is still elevated but jobs_lateness stays within a few seconds, because base nodes don't pin an entire platform triggers on one node.
Document Management PRB2056202	Replace 'Now Assist' and 'Sparkle' with 'ServiceNow Otto' in screenshots, images, and icon assets	All UI visual elements should be updated across SmartDocs, Doc to Voice, and SmartRedact to reflect the ServiceNow Otto branding changes. This includes both images/screenshots and icon assets.
Document Management Services PRB2039067	Smart redaction with redaction codes and notes for documents	This is a product update.
Document Management Services PRB2056919	SmartDocs enablement in UI16 and ServicePortals	This is a product update.
Document Viewer PRB2056915	Support in document viewer for the doc to voice agent	This is a product update.
Email Notifications PRB2036855	Agent workspace case emails display multiple 'To' names if the email recipient has multiple contact records	If a contact has more than one contact record in the system (same email address), the system displays the names of all the contacts when viewing an email sent to that person on a case.	Navigate to customer_contact table. Create multiple contact records with the same name and email address, but a different user ID. Navigate to CSM workspace. Create a case. Add one of the contacts created above as the case contact. Send an email to that contact. Open the sent email. Expected behavior: Only one contact is displayed in the To field. Actual behavior: The To field lists all the contacts with the same name and email address.
Email Notifications PRB2050616	Email client APIs shouldn't be public
Embedded Help PRB2056339	Replace all occurrences of 'Now Assist' text with 'AI' across the Help Center and embedded help		Open Help Center or embedded help. Open a document that was created using AI-generated help. Observe that 'Now Assist' text is displayed for AI-generated help articles.
Event Management PRB2052731	Remove the step in mixed grouping that allows alerts to join existing groups before the grouping definitions are run in order	The mixed grouping engine handles alerts in two steps. First, it checks whether an alert can join a group that already exists—without following the order the definitions are set up in. After, it goes through all the definitions in their proper order, where it's also allowed to create new groups. The first step disregards the configured definitions order; as a result, an alert can end up in a group created by a lower-priority rule instead of a higher-priority definition. The definition order exists specifically to control which definition gets to group an alert first, so skipping it defeats the purpose of the order. In conclusion, the two-step process should be removed. The mixed grouping engine should go through the definitions once, in order. For each definition, it should allow alerts to either join an existing group or create a new one.	Create two grouping definitions, Definition A and Definition B, with Definition A placed ahead of Definition B in the rule order. Send alerts A1 and A2 that fit Definition B. Wait for the grouping job to run. Create Group B. Send alerts A3 and A4 that fit both Definition A and Definition B. Wait for the grouping job to run. Check which group A3 and A4 end up in. Expected behavior: A3 and A4 form a new group created under Definition A, because Definition A comes first in the rule order. Actual behavior: A3 and A4 are added to the existing Group B.
Event Management PRB2053524	Azure monitor for 'Issue' integration, and UI enhancements in Express List and alert record	This is a product update.
Flow Engine PRB2059346	Flow context gets completed, but the event isn't relayed to playbook engine, so the activity context doesn't get completed		Create a playbook with one normal activity (say instruction). Add one optional activity (instruction). In runtime, add six instances of the optional activity. Start marking the optional activity complete, from the last in the list to the top. Observe that the corresponding activity context is in progress, but the flow context gets completed.
Flow Engine PRB2062691	Action Fabric flow complexity is insufficient for pricing estimation exercises	The complexity_bucket attribute contains values grouped into bucket_0_10, bucket_11_50, and likely no other buckets.	Perform MCP requests on an instance with Action Fabric telemetry. View the Action Fabric telemetry via Clickhouse or logged data. Expected behavior: The exact complexity of the MCP flows are recorded. Actual behavior: The complexity_bucket attribute contains values grouped into bucket_0_10, bucket_11_50, and likely no other buckets.
Flows (Family Channel) PRB2054449	Refactor all occurrences of 'Now Assist' in Flow Designer, Action Designer, and Flow Diagram to 'ServiceNow Otto'
Flows (Family Channel) PRB2056022	Otto directive renaming for the call skill step plugin	The plugin name is being changed from 'Now Assist Skill' to 'AI Skill' (or 'AI Skill Step'). The step name is being changed from 'Call Now Skill Step' to 'Call AI Skill Step'. Also, the plugin and step description should be updated where applicable.
Hermes (Family) PRB2026691	TCP connection establishment and SSL handshake occur every minute	TCP connection establishment and SSL handshake occur every minute even when the health check is skipped under the 150-second threshold. The unnecessary overhead should be remediated.
Hermes (Family) PRB2026695	Identify and eliminate duplicate health checks where two Hermes services point to the same Hermes cluster	Normally, it'd be expected that Glide apps would use different bootstraps. Given the special condition, when the Kafka bootstraps are the same, it should avoid the duplicate connections/healthchecks between the two services.
Horizon Component Library PRB2052114	Replace the Sparkle icon with the Otto Lottie animation in AI and loader experiences
Horizon Component Library PRB2052172	Suppress console statements so browser tools aren't flooded		View the 'Theme behavior' console output. The console output should only be logged when env = development.
Horizon Empty State Component PRB2050333	Update ai-general illustration	The sparkle illustration should use the new Otto logo instead.	Open an instance. Create an experience in UIB. Add an empty state component to the stage. In the config panel, locate the illustration property. Select AI general from the drop-down list. Expected behavior: The empty state shows the Otto logo. Actual behavior: The empty state shows sparkles.
Horizon Icon Component PRB2032851	Add the Otto icon to the library		Navigate to an instance. Create a experience in UI Builder. Add a now-button-iconic component to the stage. In the configuration panel, locate the drop-down list for the icon property. Search for 'sn-sparkmoji-logo'. Expected behavior: The new Otto logo is visible. Actual behavior: The icon is missing.
Horizon Icon Component PRB2050318	Update all Sparkle icons	Icons with the Otto logo should be shown instead.	Open an instance. Create an experience in UIB. Add a now-button-iconic component to the stage In the config panel, locate the icon property. Search for an icon. Expected behavior: Icons with the Otto logo are shown. Actual behavior: Icons with sparkles are shown.
HR Service Delivery PRB1958041	After a Zurich upgrade, LE tasks aren't getting created	During case creation, the system needs to evaluate audience, which includes the creation of lifecycleEventLogger. Inside lifecycleEventLogger, the flow checks whether there is a valid HR case. Due to the restricted query business rules, the system is blocked from accessing the target records and stops when the error is thrown. This causes the 'Evaluate Audience' result to be false and the flow skips that activity creation.
HR Service Delivery PRB2014389	RCAs are generated for 'Populate Manager Reportee Count Using Eligible Users' and 'Employee​Hub​Org​Chart​ Reportee​Util​SNC'	During nightly test case execution, two RCAs are generated for the new scripts: - sys_script_include _a302f8807873 f250f877079523d275e1 and - sysauto_script_ 33a5e72453f7 b210f2ebff c230e5e69d. It is causing test failures.
HR Service Delivery PRB2021314	The Cancel UI action in an HR case checks for 'mandatory' after the case is canceled	The Cancel UI action is invoking hr_CaseAjax.cancelAction(). In this function, state and work notes are updated from the server script. Inside the callback function, there's g_form.save(), which checks the mandatory and saves the form. If the data is already saved from the server script, the system shouldn't check the mandatory.
HR Service Delivery PRB2052205	In Australia and Zurich, RCAs generated in HR Core and HR Employee Relations need to be added back to track/HR
HR Service Delivery PRB2052632	The header_config_* fields are missing from the sn_hr_core_service table in Zurich and Australia		Open a Zurich or Australia instance with the sn_hr_core and EC Core plugins installed. Observe that the header_config_* fields are missing from the sn_hr_core_service table.
Identification and Reconciliation API PRB2028002	Fix three dynamic IRE UI messaging and wording issues and the system properties in the update folder
Identification and Reconciliation API PRB2034883	Users can add or edit identification rules for hardware's child classes while dynamic IRE is enabled	If dynamic IRE is enabled, users shouldn't be able to see or edit identification rules on the hardware hierarchy.	Enable dynamic IRE. Navigate to hardware or any of the child classes. Observe that static rules are displayed and editable, even though dynamic IRE is enabled.
Identity PRB1964703	A federated ID isn't generated for all the users in the instance	There's also instances where a federated ID isn't generated for some users even when they have a unique user_name and email.
Install Base Management Store PRB2059577	Install base items and sold products should support RAC fallback mechanism	The functions _skipFetchEntities and getQRfallbackRoles should be overridden in CSMRelationshipServiceSNC in CSMRelationshipService _InstallBaseRelatedParty to allow this fallback mechanism.
Instance Data Replication (IDR) PRB1940978	There's many IDR-DCTComparisonJobs on an instance	This is likely only a cosmetic issue and not hurting performance. Instances have at least one IDRDCTComparisonJob per node, which seems excessive. For instances with many nodes, this ends up being the majority of IDR jobs. There's only a few consumer jobs per instance (not per node).	On an instance running a version lower than Australia and configured with multiple nodes, open the sys_trigger table. Check if there's multiple IDRDCTComparisonJob records present.
Integrated Email Client PRB2051938	A Now Assist reference in Email Client must be changed to Otto	This is a product update.
Integration Hub PRB2056938	Rename the 'Now Assist for Spoke Generation API Support' Glide plugin
Integration Hub Spokes PRB2053392	Users can recover box webhook HMAC signing keys via timing oracle on JavaScript == comparison and forge events to trigger arbitrary subflows	This is a product update.
Knowledge Graph (Family) PRB2016914	Update the structure to keep the Knowledge Graph configuration file in the same folder as the released description version and dynamic configuration for the Install Server URL
Knowledge Graph (Family) PRB2052948	Implement batch insert logic for affinity jobs to improve insert latency	This is a product update.
Knowledge Management PRB2054012	Rename app-common 'Now Assist' to Otto
Knowledge Management PRB2054937	True up Store apps NAKM, KC, KCInWorkspaces and ECE	This is a product update.
Knowledge Management PRB2058947	Update versions for the base instance Knowledge Management apps to include Australia and Zurich fixes
Knowledge Management PRB2060905	Updating versions for the base instance Knowledge Management apps to include Australia and Zurich fixes	There are fixes in app-knowledge-center, app-knowledge-gen-ai, app-kb-uib, and sn-enhanced-content-editor.
List Administration PRB1963197	The Duration field displays '0 seconds' in a Core UI list when the value is null		In Zurich, navigate to any table list view that contains the Duration field. Filter by '<DURATION_TYPE_FIELD>' is 'Empty'. Review that records have '0 seconds' even if there's no value in the records (empty).
List Administration PRB2044745	Now-record-list-connected sends encodedRecord as {} instead of ''	The variables contain 'encodedRecord': {} (empty object). It's treated as a non-empty, non-nil value, so the advanced qual is evaluated against an empty 'current' and wrong/empty results are returned. This is reproducible across base instances, but works correctly in Classic UI16.	In a workspace (for example, CSM/incident), open an incident record. From the right sidebar, select Create Template. Select + to add a field. Add a reference field that uses an advanced/dynamic reference qualifier referencing the 'current' object. Open the picker ('Search for Record'). Inspect the getReferenceListLayout / now​Record​List​Connected​Reference GraphQL call. Expected behavior: EncodedRecord is '', so the advanced/dynamic qualifier is correctly skipped when no 'current' is available. This matches the UI16 template behavior. Actual behavior: The variables contain 'encodedRecord': {} (empty object). It's treated as a non-empty, non-nil value, so the advanced qual is evaluated against an empty 'current' and wrong/empty results are returned.
List Administration PRB2053781	The new AI list badge in UI16 breaks list cell content wrapping
List Administration PRB2054941	Additional tags features should be supported	This is a product enhancement.
List AI Indicators PRB2051786	The horizontal scroll for AI-tagged items in the groupedBy list is broken	When the user scrolls to the right, the field isn't visible.
List AI Indicators PRB2055051	Update the Otto indicators in seismic workspace lists	This is a product update.
List AI Indicators PRB2055052	Update the Otto indicators in UI16 lists	This is a product update.
MID Server PRB2063491	'LinkedHashMap$Entry' objects connected to the LRU take a couple MB	These objects are from 'ecc_​queue_​authorization_​policy'.​	Open a local instance. Connect to the MID Server. Create a heap dump. Notice that entries of 'LinkedHashMap$Entry' for ecc_queue_authorization_policy have over 3MB for each entry.
Mobile Platform PRB1965358	Scripted screen questionnaire dependencies between fields stop working because of irrelevant evaluations using toString()	In the Now Agent app, the mandatory questionnaire doesn't honor dependency settings after an instance upgrade.	Open an instance. Impersonate 'Persona field service officer'. Navigate to the 'My work' navigation tab at the bottom. Select a work order task. Select Visit Report. Scroll down, then select Visit Report Install. Expected behavior: 'Age of responsible adult' and 'Responsible adult relationship' should be visible when 'Is the subject a child' is set to 'Yes' and 'Parent, guardian and other P/G/O' is set to 'Other'. Actual behavior: The fields show 'Age of responsible adult' and 'Responsible adult relationship' on load.
Multi-Instance Framework - Core PRB2062210	Implement Glide family changes to support discovering Hermes across data centers	This is a product update.
Multi-Instance Framework - Core PRB2063884	MIF Hermes doesn't refresh the cluster configuration when the local hermes_cluster_config has no primary or after a datacenter-rule change, causing stale/failed cluster resolution for remote owners	When instance A sends a MIF async message to instance B, it needs B's Hermes cluster details (datacenter + Kafka bootstrap servers). A keeps a saved copy in the hermes_cluster_config table and reads it in Hermes​Producer​Client .​get​Cluster​Info​Set.​ That method only calls B's live endpoint (/api/now/hermes_cluster_info, tier-2) when A has no saved rows for B. Two gaps result: First, saved rows but no primary — if A has rows for B where no one is is_primary_for_service=true (scenario like only a single non-primary cluster row), the method does not refresh. And method ensure​Topic​Location​For​Instance(String owningInstance) then finds primaryCluster == null and the send fails with 'No primary cluster found'. Second, stale rows after a datacenter change — if a MIMIR rule change moves B's cluster to a new DC, A's saved rows are outdated but still look complete (they have a primary), so tier-1 returns them and A never re-discovers → messages Navigate to the old cluster.	Verify that Instance A has a saved Hermes cluster rows for instance B (service MIF-Hermes) pointing to B's old datacenter, or a single row that isn't marked primary. Send a MIF async message from A to B. Expected behavior: A resolves B's current cluster details and sends to the correct datacenter. Actual behavior: A uses the old/incomplete saved config and sends to the wrong datacenter, or fails with 'No primary cluster found'.
Multimodal Service (Family Channel) PRB2051440	On-demand MMS submission fails	The issue occurs on instances where the multimodal service URL is configured via sys_service_endpoint instead of the system property.	Open an instance with the MMS plugin active. Leave glide.​platform_​mm_​service.​service_​url empty. Configure the Multimodal service endpoint via an active sys_service_endpoint record. Confirm that the batch/scheduled MMS job reaches the service. Observe that it resolves the URL from sys_service_endpoint. Call new MultimodalServiceUtil() .​submit​On​Demand (attachment​Sys​Id,​ query). Observe that the on-demand MMS submission fails.
Next Experience Unified Navigation PRB2053821	Update to Otto branding in Next Experience	This is a product update.
Now Assist in Virtual Agent PRB2056609	Change the Knowledge Graph (KG) defaults in NAVA and Now Assist Portal	Today, the KG default is User NLQ graph. That should be changed so the defaults are: In Now Assist Virtual Agent for natural language query, change the graph to Enterprise Graph (Small) and select the tag as 'VIRTUAL AGENT DEFAULT TAG'. In Now Assist Panel for natural language query, change the graph to Enterprise Graph (Small) and select the tag as 'NOW ASSIST PANEL DEFAULT TAG'.
Now Assist Nextwave Experience PRB2059502	Add sys_props for caching in sys_og_conversational _cache_configuration	If props are enabled on the instance, that should be reflected via cache service instantly.
Now Assist Nextwave Experience PRB2061245	SessionController shouldn't try to refresh AuthorizationInfoBundle, as this is done in the Glide handshake process	When a topic execution is initiated, the conversation ID isn't available in certain Glide log entries' context maps, which complicates troubleshooting across different trace IDs instead of a unified conversation ID. The conversation ID should be available in the cache and topic execution context map in the syslog table.
Now User Experience PRB1960854	JavaScriptCommentStripper doesn't work properly with the upgraded highchart bundle (12.3.0)	There's a runtime exception: 'org.​openqa.​selenium. ​Javascript​Exception:​ Document was unloaded'.	Run IT test​Outage​Association​With​Case​IT .​check​Child​Case​Linked​To​Parent​Outage.​ Observe that it fails on main.
OAuth 2.0 integration PRB2032073	Chat-Input is available to requester when conversation ends
OAuth PRB2052671	Support to create an OAuth entity record within the caller scope	This is a product update.
On-Call Scheduling PRB1975559	On-call contact information isn't visible for some users for some attempts	This issue only occurs when logged in as a user.
On-Call Scheduling PRB2034496	In a SMS action in the subflow 'On-Call', the 'Check Assignment Response' doesn't use Notify	The 'On‑Call: Check Assignment Response' flow uses the legacy SMS send action instead of the Notify‑based SMS action.
On-Call Scheduling PRB2039960	The NotifyUtils check for notification devices should return empty when the ACL check is successful
OneExtend PRB2041342	AI Summary card stays in perpetual loading state	This issue occurs for some Business Application records using the Record Summarization capability. There are no LLM calls and no errors.	Navigate to APM Workspace. Open a Business Application record. Observe that the AI Summary card stays in perpetual loading. Check sys_generative_ai_log. Observe that there's no LLM call logged for this request. Check syslog. Observe that there's no errors; the full pipeline completes with cache HIT. To compare with a working BA record, navigate to service-​now.​com/​now/​apm/​application-​rationalization.​ Select the Now Assist icon on business apps. Observe that the summary is generated. It generates LLM calls and renders correctly.
OneExtend PRB2056745	Guardian preprocess flow resolves getGeoRoutingDetails() multiple times per request in NowLLMIntegration GuardianProvider	NowLLMIntegration GuardianProvider. shouldUseGatewayService() and addLLMGatewayRoutingHeader() each independently call through to GeoRoutingServiceImpl .getGeoRoutingDetails() > resolveGeoRoutingDetails(). ShouldUseGatewayService() itself is invoked from multiple call sites across a single request's lifecycle (transformRequest, generateTrustBuilderInputs, tryLogTrustBuilderResults, and twice within getUrl()). None of these calls are memoized, so resolveGeoRoutingDetails() re-executes its full resolution logic (potentially including the licensing entitlement API call) on every invocation within the same request, even though the underlying geo-routing state can't change MID-request.	Trigger a Guardian moderation request that routes through NowLLMIntegration GuardianProvider (LLM_GENERIC_ SMALL_MODERATIONS model). Trace/log calls into GeoRoutingServiceImpl .resolveGeoRoutingDetails() (or set a breakpoint) during a single request's transformRequest()/getUrl() lifecycle. Observe that resolveGeoRoutingDetails() executes repeatedly (up to six times found via code trace) instead of once per request. When the 0$ SKU entitlement isn't active, each of these calls re-invokes the expensive isEntitlementActive WithLicensingAPI() licensing call, since resolveGeoRoutingDetails() has no per-request memoization. Only the underlying getGeoRoutings() /getGeoRoutingConfigs() cache calls are cached via ADomainAwareCache.
OneExtend PRB2057127	Multiple GenAI logs are created in skill chaining execution	OneExtend Execute and ExecuteSecure are missing the offGlideChainingEnabled flag.	Execute a skill chaining. Observe that the Generative AI logs are created two times.
OneExtend PRB2059503 KB3142021	Summarization records aren't displaying a proper response	Incident, Change, and Case summarizations are failing with errors when users select the Summarize button: 'Summarization could not be completed because access to the base table Case was unsuccessful'. Error logs: 'Error sending to unified_​short_​url_​active_​ 1.​.​.​Status 500 - [Internal Server Error] [{'success':​false,​'status​Code':​429,​'message':​'Too many concurrent data insert operations in progress - additional rebuild request being ignored',​'timestamp':​'2026-​07-​17​T07:​00:​17.​752637494​Z',​'results':​{}​}​]​'.​	Refer to the listed KB article for details.
Performance Analytics PRB1899019	PA scores aren't retrieved immediately after enabling data snapshots	The 'No scores available' message is displayed. It remains the same until mining is complete, which shouldn't happen.	Provision an instance with the com.snc.pa.premium and com.snc.pa.mlb plugins installed. Navigate to the number of open incidents. Make sure there isn't a data snapshot already created for the incident table. Remove the age breakdown. Run the classic Historic PA Job to view scores. Enable data snapshots for this indicator. View the scores. Observe that the 'No scores available' message is displayed.
Performance Analytics PRB2032037	The Platform Analytics dashboard filters ignore Element Security Lists and display 'Nothing is available'	If the user sets up an Element Security List for a breakdown source that's used as a filter in a Platform Analytics dashboard, the filter shows 'Nothing is available'. This happens even though the user has valid access to the dashboard and the underlying data.
Platform Analytics Dashboard API PRB2037426	scoreType is forced to 'latest' for non-aggregate indicators due to an undefined aggregateIndicator check	When a visualization widget uses a classic indicator (sourceType: 'indicator') with applyDateRange enabled, the scoreType property is incorrectly forced to 'latest' even when the metric is not an aggregate indicator. This happens because the aggregateIndicator check in the scripted policy evaluates 'undefined !== ''' as 'true', even though the metric object doesn't have an 'aggregateIndicator' property at all. The user loses the ability to select other score types (sum, average, etc.) for non-aggregate indicators, and any previously configured scoreType value (for example, 'sum') is overwritten to 'latest'.	Add a single-score (or any visualization that uses the shared date-range policy) widget to a dashboard. Configure it with a classic PA indicator data source (sourceType: 'indicator') that isn't an aggregate indicator (for example, 'Number of new requested items'). Enable 'Apply date range' (applyDateRange: true). Set scoreType to 'sum' (or any value other than 'latest'). Open the widget configuration again. Expected behavior: The scoreType remains 'sum' and all score type options are available in the drop-down list. Actual behavior: The scoreType is forced to 'latest' and the drop-down list is inactive.
Platform Analytics Dashboard API PRB2038900	Platform Analytics Release (PAR) dashboard's saved filters drop on 'GET' (200, empty filters) when a referenced saved/library filter is unresolvable (deleted or cross-domain)	Filters​Service.​remove​Invalid​Filter (inline path) had a guard that returned '' whenever availableFilters.size() didn't match the layout's filter-component count. getAvailableFilters can't resolve a stored_component whose par_component_filter record is deleted or domain-invisible, so a single unresolvable reference caused all of the dashboard's saved filters to be discarded for every user. The PUT path does no validation and saves correctly; only the GET-time validation drops them. On domain-separated (MSP) instances, this is easily hit because dashboards in user sub-domains reference global saved filters.
Platform Analytics Dashboard API PRB2039418	List migration logic should be added for the highlighted values config	When the user upgrades, the corresponding properties should be migrated to the new list visualization.	On a Yokohama instance, create a list-simple visualization. Enable fetchHighlightedValues. Add some highlightedvalueconfigid to the list-simple visualization. Upgrade to Zurich, Australia, or Brazil. Observe that the corresponding properties are migrated to the new list visualization.
Platform Analytics Dashboard API PRB2050660	Par_dashboard_tab records are created unintentionally when a non-admin user accesses an inactive dashboard	When a non-admin user accesses an inactive (deactivated) Platform Analytics dashboard through the workspace URL, one or two new par_dashboard_tab records are created unintentionally.	Impersonate a user. Navigate to Platform Analytics > Library > Dashboards. Create a dashboard without any dashboard tabs. Open the dashboard record using a URL like 'https:​/​/​<instance-​id>​.​service-​now.​com/​ par_​dashboard.​do?​sys_​id=​<sys_​id>​'.​ Clear the Active option. Update the record. As the same user as step 1, access the dashboard via a URL like 'https:​/​/​<instance-​id>​.​service-​now.​com/​ now/​platform-​analytics-​workspace/​dashboards/ ​sys-​id/​<sys_​id>​'.​ Check par_dashboard_tab. Expected behavior: The par_dashboard_tab doesn't have a record for the test dashboard. Actual behavior: Once the user accesses the inactive dashboard, one or two new par_dashboard_tab records are created unintentionally.
Platform Analytics Filters PRB1968556	Users are unable to filter on the sys_class_name column of the kb_knowledge table in a new Platform Analytics dashboard	The column 'Class' isn't present.	Navigate to self-service dashboards (also known as Core UI dashboards). Create a dashboard. Add an interactive filter to the dashboard that filters on table 'kb_knowledge' and column 'Class'. Observe that it works. Save the dashboard. Migrate it to Platform Analytics. Observe that the filter works. Create a Platform Analytics dashboard. Add a filter. Select table 'kb_knowledge' as the source. Try to select column 'Class'. Expected behavior: The user can choose the column 'Class'. Actual behavior: The column 'Class' isn't present.
Platform Analytics Filters PRB1993721	A dashboard's Year to Date (YTD) filter fails to apply on a list unless it's reset	A dashboard's YTD filter doesn't consistently apply the correct date range to list widgets, resulting in incorrect list results until the YTD filter is manually cleared or reset.
Platform Analytics Filters PRB2018497	When using a date/time filter, the 'Today' predefined range incorrectly starts from the current hour instead of all of today	When setting up a date filter in Platform Analytics, if the predefined range of 'Today' is added and 'Allow time selection' is set, it will default to the start time of Current Date + Current time, and the end time of Current Date + 23:59. This means it is not showing the whole day, and instead just starting from the current moment and going into the future (and therefore is unlikely to show any records).	Create a Dashboard in Platform Analytics. Add a date filter. Add the 'Today' predefined range. Add any other predefined range for easy testing. Set the default value to 'Custom Range'. Enable Allow time selections. Save changes and Navigate to the date filter. Select any predefined range other than 'Today'. Select Today. Notice that the date will update correctly, but the start time will update to the current time.
Platform Analytics Filters PRB2021606	An interactive filter can be cleared via a double-click even when the 'Allow User to Clear Filter' option is turned off	Users are able to clear the applied interactive filter by double-clicking the selected filter value even when the 'Allow user to clear filter' option is turned off in the filter configuration and clear filter option isn't appearing. Additionally, when multiple interactive filters are configured with filter interactions/dependencies, the dependent filters still display the option to clear the filter. This behavior creates inconsistency with the expected functionality of the 'Allow user to clear filter' setting, as users are still able to clear filters through alternate interactions.	Create a dashboard. Add a visualization with an interactive filter. Configure the interactive filter to turn off the option 'Allow user to clear filter'. Apply/select a filter value on the dashboard. Observe that there's no 'Clear filter' option. Double-click the selected filter value. Observe that the filter is cleared even though the clear option is turned off. Configure another interactive/dependent filter. Observe that the dependent filter still displays the option to clear the filter. Expected behavior: When the 'Allow user to clear filter' option is turned off, users shouldn't be able to clear filters through any UI interaction, including double-click actions or dependent filter interactions. Actual behavior: Filters can still be cleared through double-click interaction and dependent filters display clear filter options despite the setting being turned off.
Platform Analytics Migration API PRB2034116	Upgrading from Yokohama to Zurich causes all pa_widgets to be migrated after upgrade	The pa_widgets table doesn't have an active field.
Platform Analytics Migration API PRB2037846	The Migration Center summary count doesn't match the list count for fully migrated dashboards	The Get Migration Summary Scripted API needs to be updated so that it calculates the total number of migrated dashboards in the same way as the Migrated List.	Scenario 1: Create a few Core UI dashboards. Migrate them. Navigate to the PAR Dashboards table. Delete one or more migrated dashboards. Observe that the Migrated List count and the Summary count don't match. Scenario 2: Create a Core UI dashboard containing a Dynamic Content widget. Migrate the dashboards. Navigate to the par_coreui_migration _bridge_dashboard table. Remove the record(s) associated with the dashboard that contains the Dynamic Content widget. Observe that the Migrated List count and the Summary count don't match.
Playbooks (Family Channel) PRB2025987	A user can't view a playbook without record create access
Playbooks (Family Channel) PRB2052473	The getPlaybookContextParentRecord API should be exposed as scriptable	The Playbook MCP tool code doesn't use the existing API get​Playbook​Contexts​By​Parent​Record to launch and continue playbook execution. This API internally handles all validation and permission access, which results in duplicate logic. The current java API should be exposed as a scriptable API to use at Playbook MCP tool.	Add any compatible playbook as MCP tool. Try to launch playbook through the Claude client. Observe that it launches playbook, but it uses custom logic to validate access and permission.
Playbooks (Family Channel) PRB2052670	'Completed By' and 'Completed On' for Activity Contexts	This is a product enhancement.
PPM Standard PRB2037975	Relax RIDAC table business rules	This is a product update.
Process Mining PRB1957306	Scheduled tasks are shown in the 'Pending' state after the execution tracker is deleted	The scheduled task shows up on workbench with the 'Pending' state and no actions can be taken on it.	Create a scheduled task. Delete the execution tracker associated with the task. Observe that the scheduled task shows up on workbench with the 'Pending' state. No actions can be taken on it.
Process Mining PRB2022864	Users are unable to edit access for Now Assist creator's 'Work Notes Analysis' skill		Navigate to a Zurich instance. Impersonate any admin user. Navigate to Now Assist Admin. Open Creator > Skill > Work Notes Analysis. Try to add an extra role beside the given role. Expected behavior: The user should have edit permission to edit. Actual behavior: The user is unable to edit permissions.
Process Mining PRB2035128	Meter-based guardrails and controls	This is a product enhancement.
Process Mining Workspace PRB2022151	The record count for child tables shows the total records and ignores any applied filter	Since this is expensive to calculate, the recommendation is to hide that counter for child tables altogether.	Create a project. Add a child entity. Add a filter to the child entity. Select the button to show total number of records for that child entity. Expected behavior: Only the records in scope are shown. Actual behavior: The total number of records in the table are shown.
Process Mining Workspace PRB2023836	The Process Mining usage table (promin_metered_usage_data) must be preserved across clones	The Process Mining usage table (promin_metered_usage_data) is overwritten during instance clone operations, causing mined record counts to be duplicated and incorrectly reported when users mine on non-production after cloning from production. Process Mining licensing counts unique mined records per instance (not across instances).	Mine records in production. Clone production into a non-production instance. Mine again in non-production Expected behavior: After any clone operation, the Process Mining usage table on the target (cloned) instance is reset to its original (empty/baseline) state, so usage counting starts fresh/continues per instance. The configuration that enforces this should be mandatory on every clone and read-only / protected. Actual behavior: The cloned non-prod instance inherits production's usage history, but subsequent mining activity overwrites or duplicates entries. This breaks per-instance usage reporting and creates a double-charging risk.
Process Mining Workspace PRB2024798	Summary top banner metrics isn't displaying until the project is reopened or the page is refreshed
Process Mining Workspace PRB2034613	Users can't edit access for the Now Assist creator 'Work Notes Analysis' skill		Open a Zurich instance. Impersonate any admin user. Navigate to Now Assist Admin. Open Creator > Skill - Work Notes Analysis. Try to add an extra role besides the given role. Expected behavior: The user has permission to edit the field. Also, the workbench should respect the permission set at skill level. Actual behavior: The user is unable to edit permissions.
Process Mining Workspace PRB2038768	Scheduled tasks are shown in the 'Queued' state after the execution tracker is deleted	Scheduled tasks show up on workbench with the 'Queued' state, and no actions can be taken on them.	Create a scheduled task. Delete the execution tracker associated with the task. Observe that the scheduled task shows up on workbench with the 'Queued' state and no actions can be taken on it.
Process Mining Workspace PRB2050830	Wipe out promin_metered_usage_data across non-prod and rely on the Glide non-prod/prod property, not on the Open/Closed property	When users clone production to a non-prod instance, the non-prod promin_metered_usage_data table is wiped and replaced with prod's data, so non-prod reflects a copy of prod rather than real usage. When users clone prod to several non-prods at once, each carries prod's usage table. Subscription Management's flat aggregation then counts the same prod records once per instance, so reported usage scales linearly with the number of non-prods a user maintains, independent of actual mining behavior.
Process Mining Workspace PRB2052878	Otto directive for process mining	The Otto rebrand should be completed for work notes analysis, intent and activity analysis, process highlights, generate process config with AI, and playbook mining.
Process Mining Workspace PRB2055568	In Zurich, a false 'access denied' error appears when saving table/filter changes on an existing project	Process Mining table-access validation incorrectly denies save actions for users who have genuine read access to the table via a conditional ACL. The system throws: 'You do not have access to {table} table'. This happens because it checks permissions against an empty test record instead of real data. The error is a false negative in the platform's own pre-save validation, not an actual permissions gap. Users with access via an unconditional ACL don't receive the error.	Create a dedicated test role (for example, test_incident_reader). On the incident table, add a new read ACL. Set the role to the new test role. Add a data condition (for example, assignment group is <specific group>). Create a test user with only that new role (make sure there's no other role tied to an existing unconditional ACL on incident, like itil or sn_incident_read). Create an incident record matching the condition (for example, the same assignment group). Impersonate the test user. Confirm via the incident list view that the user sees only that one matching record. As that user, open Process Mining. Create or open a project with a table configuration on incident. In the 'Scope your analysis' step, edit a filter. Save. Expected behavior: The save succeeds. Actual behavior: The message 'You do not have access to incident table' appears, despite step 8 confirming real access.
Project Management PRB2006929	Deleting a project task with MS Project creates an orphaned resource request		Log in to any instance. Open the new Project Workspace. Select the New button to create a project. Select the Menu button (...) > Import from MS project. Upload a MS Project schedule. Create a resource request for Project Tasking. Navigate to the resource module as the resource manager and approve the task. As the project manager, delete the task in the MS Project schedule. Return to the project record. Re-upload the updated MS Project. Expected behavior: The re-upload of the schedule isn't permitted because of the related record. If it does allow the re-upload, it should cascade the delete to the related records. Actual behavior: The resource manager can still see the resource request, but the task name is blank.
Project Management PRB2023802	The PowerPoint export service loses the connection	The export functionality is broken. When using the export feature, it displays 'Unable to complete request'. There are no log entries available for this error message. In order to exclude issues with the built-up service, there are regular pings every 15 minutes from the instance to the backend service, but those pings stop at the end of the day. After manually updating the entry in sys_service_endpoint, the pings start and exporting to PowerPoint is available.
Reporting PRB2040639	Require authentication on a scripted rest API reporting_alias
Roles PRB2052882	Inherited roles aren't added back when patcher is on and the state of 'User has role' is changed from pending_approval to active	When the user updates the state of the 'User has role' record to active, no records are created for inherited roles.	Enable the glide.​security.​inh_​count _​patcher.​enabled property to true. Assign a user with an admin role and with the state as pending approval. No 'User has role' records should be created for inherited roles. Update the state of the 'User has role' record to active. Expected behavior: 'User has role' records are created for inherited roles with the state as active. Current behavior: No 'User has role' records are created for inherited roles.
Scheduled Jobs PRB2015240	Create a guard rail to protect the scheduler against misconfigured nodes to avoid impact to critical jobs	Misconfigured nodes can severely impact the job scheduler, causing business critical jobs to not get executed.	On a two (or more) node cluster, configure one of the nodes to have schedulers=any and participation=standby. Create 1000 pinned jobs with priority=100 and pin to that node. Wait a few minutes. Observe that both this node's and other nodes' pinned jobs with priority>=100 no longer execute.
Schedule Optimization (Glide Family Channel) PRB2034899	Intraday optimization is stuck in 'In progress'	Intraday optimization jobs may become permanently stuck in the 'In Progress' status due to a timing issue in how solution metadata is processed internally, which can result in job status updates being lost.	Configure Schedule Optimization with intraday optimization enabled with territory model. Set up an intraday optimization schedule with multiple overlapping territories. Trigger an intraday optimization run. Check wm_intraday_job_m2m_qualifier. Observe that some of the records remain stuck in in_progress status indefinitely, even after all the solutions are processed or the records show no processed task counts, total task counts, and/or progress bar. During normal intraday runs with concurrent solution processing, the race condition between SolutionMetadataProcessor and SolutionProcessorSNC can cause lost updates on ml_solution.solution_metadata, also resulting in stuck in_progress records.
Schedule Optimization (Glide Family Channel) PRB2035210	When work schedules aren't created in WFO, GraphQL fails a few times		Create an empty work schedule in WFO. Send a QualifierData GraphQL call. Observe that GraphQL fails a few times.
Schedule Optimization (Glide Family Channel) PRB2041460	In Zurich, an intraday job takes around 12 minutes to be submitted to ML Scheduler		Log in to a Zurich instance. Run an intraday job with 200 qualifiers. Observe that it takes around 12 minutes to be submitted to ML Scheduler.
Schedule Optimization (Glide Family Channel) PRB2051159	Intraday optimization is stuck in 'In progress'	The race condition is in solution processing. Glide mutex API needs to be added in global scope.	Configure Schedule Optimization with intraday optimization enabled with territory model. Set up an intraday optimization schedule with multiple overlapping territories. Trigger an intraday optimization run. Check wm_intraday_job_m2m_qualifier. Observe that some of the records remain stuck in the in_progress status indefinitely, even after all the solutions are processed or the records show no processed task counts, total task counts, and/or the progress bar. During normal intraday runs with concurrent solution processing, the race condition between SolutionMetadataProcessor and SolutionProcessorSNC can cause lost updates on ml_solution.solution_metadata, also resulting in stuck in_progress records.
Schedule Optimization (Glide Family Channel) PRB2056649	There's a double booking issue in conflict resolution		Configure Schedule Optimization with intraday optimization enabled with territory model/assignment group. Set up an intraday optimization schedule with multiple overlapping territories/assignment groups. Trigger an intraday optimization run. Open dispatcher workspace. Observe that a few work order tasks are double booked.
Schedule Optimization (Glide Family Channel) PRB2057489	SO conflict resolution is unassigning locked tasks	A conflict resolution fix was implemented to unassign tasks from the solution when the assignee has a conflict. However, it unassigns tasks even if they were locked after solution processing began. There should be a filter to skip the unassignment if the task has transitioned to a locked state.
Server-side scripts PRB2051345 KB3128196	Automatically created KittyScript exemptions are incorrect when using 'new GlideRecord'	When GuardedScript/KittyScript checks scripts in the exemption table, it does by checking a _normalized_ version of the script, where all literals are removed to ensure that it deduplicates similar scripts where the only difference is (say) a 'sys_id'. This normalization does _not_ apply to the constructor of 'GlideRecord' where it doesn't want a blanket exception for _all_ tables.	Refer to the listed KB article for details.
Service Catalog Portal Widgets PRB2054951	ServiceNow Otto rebranding for catalog item slot-fill	This is a product update.
Service Catalog PRB2054931	ServiceNow Otto rebranding for catalog item generation with Text2Catalog family support	This is a product update.
Service Level Management PRB2001036 KB3137219	In SLACalculatorNG, an unbounded schedule/DurationCalculator cache causes memory growth during bulk SLA recalculation	During bulk SLA recalculation, SLACalculatorNG accumulates one GlideSchedule and one DurationCalculator instance per unique schedule+time zone combination into unbounded object maps (this.schedules and this.durationCalculators). These maps are never evicted and grow proportionally to the number of distinct schedule/time zone pairs encountered across all task_sla records in the query. In instances with many SLA definitions backed by different schedules, this results in unbounded heap growth for the lifetime of the bulk calculation, degrading instance performance and potentially causing out-of-memory conditions under load.	Refer to the listed KB article for details.
Service Mapping PRB2036299	After upgrading to Australia, parent.process.pid fails to resolve, which causes the 'Get Process' operation to return the full process list instead of the targeted parent process	The user can run a customized Tomcat WAR pattern for Service Mapping discovery. Following the Australia upgrade, step two of the 'Identification' section stops working during Service Mapping discovery. The step uses a 'Get Process' operation with process_id = get_attr {'parent.process.pid'} to locate the parent Windchill Java process. This expression fails to resolve, causing the step to run with no PID filter and return the entire process list on the host instead of the targeted parent process.	Upgrade to the Australia release. Run discovery on the service map with Tomcat CI. Observe that the 'Get Process' operation returns the full process list from the host instead of the targeted parent process.
ServiceNow SDK (Glide) PRB2013533	There's an unhelpful error when installing an app with an invalid scope prefix	Error: '[now-sdk] ERROR: Exception occurred while installing application/Unable to install application as application was null. Error: Exception occurred while installing application/nUnable to install application as application was null'.	Create an app with now-sdk init without a valid auth saved. The resulting app has an x_ prefix and gives an error when trying to install on an instance.
ServiceNow SDK (Glide) PRB2033218	Fluent App metadata deletions aren't propagated when installing from the app repo	The issue occurs because of missing sysmetadatadelete records.	Delete metadata from Fluent App. Build and deploy to the instance. Publish to the app repo from the instance. Install it on a different instance that doesn't contain the delete from app-repo. Expected behavior: The delete is applied and the metadata is removed. Actual behavior: The metadata is not removed.
ServiceNow Studio (Family Channel) PRB2063560	True up the Glider Store app
Service Portal PRB2056735	Update the accessibility voice-input toggle label in Service Portal for the Otto rename	As part of the 'Now Assist to Otto' rename, the accessibility toggle label must be updated in Service Portal. It's currently 'Enable voice input for Now Assist in Virtual Agent'. The new value will be similar to 'Enable voice input for ServiceNow Otto'.
Service Portal PRB2056737	Portal-core widget UI and theming for new Otto onboarding modal in the Service Portal	The modal introduces users to the ServiceNow Otto experience on first login. Because Lit-based workspace components can't be embedded in angular portals, a dedicated Service Portal implementation is required. The widget lives in the platform layer and is shared across all portals. No per-portal or user uptake is required.
Service Portal PRB2056743	Eligibility and admin control for new Otto onboarding modal
Sidebar (Family Release) PRB2058602	Typing indicator no longer works		Open two browsers. Log in as two different users. Create a sidebar chat for the two users. Start typing. Observe that there is no typing indicator.
Software Asset Reconciliation PRB2000893	Recon updates are dropped by a database, causing license metric results (LMR) and product results to not update
Software Lifecycles PRB2004396	A calculated lifecycle should consider other phase start dates before creation
Software Lifecycles PRB2010294	Matching a common platform enumeration (CPE) to DMs should ignore CPEs that are version less	sn_itam_samp. MatchVulnerableSoftware ToDiscoveryModels should ignore CPEs that are a version less. Vulnerabilities are reported against versions of a product, but due to National Vulnerability Database (NVD) data or corrupted NVD data, CPEs may be a version less.
Software Lifecycles PRB2012476	Matching a common platform enumeration (CPE) to DMs should ignore CPEs edition and software edition	There are two Edition fields on CPE, 'Edition' and 'Software Edition'. It queries against both, but 'Edition' is a legacy and deprecated field from the old 2.2 standard. 'Software Edition' was introduced in 2.3.
Standard Ticket Page PRB2054934	ServiceNow Otto rebranding for Now Assist in 'Standard Ticket' page family	This is a product update.
Stream Connect Core PRB2019118	Generative AI logs are missing
System Events PRB2003217	Fix behavior of event processing tables during the clone operation and clean up invalid records/jobs during the clean up script	During cloning on Zurich instances, events processing framework tables' configuration aren't preserved, causing errors in transaction logs due to orphan records for events jobs. An orphan record occurs when a sys_trigger record exists for an event processing job, but the corresponding entry in the sys_processing_framework_job table is missing. This causes the ProcessingFrameworkJob to fail when attempting to retrieve the scheduled job context. Log example: 'ProcessingFrameworkJob SEVERE ProcessingFramework > Failed to get the schedule job context'.
System Events PRB2007066	Check to correct Legacy and Delegated Flow Engine jobs that are created during Upgrade	This is a product update.
System Import Sets PRB2038233	JDBC connections are left dangling and unclosed	Every time the user issues a change credential signal to MID during a JDBC import, the number of open connections grows.	Set up a MID server with a long running JDBC import that can be run through it. Start an import. While it's running, issue a change credentials signal to MID using the 'Refresh Credentials' related link on the MID server record. Repeat. Check the database for open connections from MID. Observe that the number of open connections grows each time.
System Web Services PRB2029605	There's hourly recurring OAuth authentication failures on an outbound REST from ST ServiceNow to CG ServiceNow: 'User is not authenticated. OAuth token has expired or has not been retrieved'	The reported issue affects the outbound integration between ServiceNow and the external ServiceNow instance through the iPaaS layer using OAuth 2.0 Client Credentials authentication. The integration fails specifically during the final one-minute window before the OAuth access token expiration. During normal token validity, the integration works successfully. Starting one minute before token expiration, outbound calls from ServiceNow fail. Failures occur in: REST Message, REST API Step, Flow Designer executions, and custom scripts. Once the token fully expires, ServiceNow successfully retrieves a new token and the integration resumes functioning normally.
Territory Planning PRB2039182	Updating potential territories for a task doesn't work as expected	The Business Rule that updates the potential territories is currently triggered only when the location changes. It doesn't consider the consider_​potential_​territories _for_schedule_optimization boolean field.	Create a work order task. Change the location to see the eligible territories as per the new changed location. Expected behavior: The Business Rule that updates the potential territories should consider the consider_potential_territories _for_schedule_optimization boolean field. Actual behavior: The Business Rule that updates the potential territories is currently triggered only when the location changes and doesn't consider the consider_potential_territories _for_schedule_optimization boolean field.
Time-Limited User Roles PRB1972197	When a user is logged in, updating an existing record with a new end time in the 'Time limited user roles' table causes never ending information messages on all pages		Open a Zurich instance. In a separate browser or incognito window, log in with a user that has an 'itil' role. In another window, log in with a user who has an 'admin' role. With the admin user, navigate to All > User administration > Time-limited roles. Create a record that grants an 'admin' role to the user. With the user that had only an 'itil' role but received the 'Time limited admin' role, open an incident list and then a record. Expected behavior: Users see an information message once. Actual behavior: Users see a message about roles that are granted or removed, but closing them isn't helping.
Trace Collector - Family Release PRB2056896	Trace collector MID code	This is a product update.
Trace Collector - Family Release PRB2059215	Azure classic trace collector doesn't consider Credential IDs for AI and ML services	AzureTraceCollector ignores credential IDs supplied as config update parameters. It only considers credential alias names.
UI Actions PRB2051051	The field error message isn't shown when the error comes from the backend via business rule abortAction	The error message isn't shown for the short_description field in Australia. However, it's visible in Faster or PAXE Integration.	Create an Insert/Update business rule with abortAction (for example, current.​short_​description.​set​Error('Error!')​;​ current.​set​Abort​Action(true)​;​)​.​ Create an incident. Fill in the mandatory fields. Attempt to save it. Observe that the error message isn't shown for the short_description field.
UI Field Administration PRB2054948	ServiceNow Otto rebranding for AI indicators, AI indicator pop-overs, and Task intelligence prediction	This is a product update.
UI Field Administration PRB2054949	IndexName GraphQL component and choice fallback	This is a product update.
UI Field Administration PRB2054953	Inject '*' options into composite_name table and field picker choice lists	This is a product update.
UI Form Administration PRB2019275	There's layout and scrollbar flickering when entering info, internal and external, in HTML Journal fields	There's layout and scrollbar flickering when entering info, internal and external, in HTML Journal fields, caused by automatic browser window resizing. The issue occurs randomly and is not related to a specific case or text size.	Change the glide.ui.journal.use_html system property to true. Open any incident in any workspace. Type multiple lines in internal info or copy/paste some text until the scroll bar appears. Remove enough lines so that it's one new line away from displaying the scroll bar. Type some words randomly to add enough words to create a new line. Observe the flickering issue.
UI Form Administration PRB2033143 KB3139369	The 'Preview this record' icon isn't working in the Safari browser	Selects on the button components and controls aren't working in Safari for UI16. This issue is only reproducible in the Safari browser. Non-Safari browsers aren't impacted and they don't face this issue.	Refer to the listed KB article for details.
UI Form Administration PRB2054945	Otto brand change in form notification messages	This is a product update.
UI Form Administration PRB2056498	The parentRecordSysId is missing from ORM Workspace payload template, so the getFilterQuery API breaks	The getFilterQuery REST API (/api/now/ related_list_item_filter/ getFilterQuery) is being updated to require four mandatory parameters: tableName, parentFieldName, parentRecordSysId, and referencedFieldName. The Configurable Workspace for Order Management (sn_app_orm_wksp) payload definition doesn't include parentRecordSysId in its payload template. Once the change occurs, any UI action using the payload receives a 400 error from the API.	Provision an instance with the plugin com.sn_app_orm_wksp installed. Activate the plugin. Trigger any UI action that uses the payload definition for event mappings. Observe that the call to /api/now/ related_list_item_filter/ getFilterQuery returns {'error': 'Invalid inputs'}.
Upgrade Center PRB1950446	The 'Flow Designer' module directs to 'Page Not Found' after an instance upgrade on some nodes	The Process Automation > Flow Designer module redirects to 'Page Not Found' on some nodes after an instance upgrade. The main issue is that nodes tried to download app packages from Store, but Store responded with a 400 error because the platform version for the instance isn't updated on the Store end yet, and the request is considered incompatible. However, during an upgrade, it shouldn't request app packages from Store.
UX Framework PRB2026005	UXF should let the inbox know when the workspace is not the active experience	Currently, when the agent navigates from workspace to home, the inbox is still present in the DOM. Because of this, agents are still assigned work items even when they aren't in the workspace.
UX Framework PRB2035397	After typing in an input field on record page load, the focus shifts intermittently	After a few words, the focus shifts to the 'Tags' section and it continues to write in that section.	Navigate to CSM/FSM workspace. Open the list module. Navigate to 'All Cases'. Open any case form with tags displayed on it. Start typing in an input field (for example, Compose comments). Expected behavior: The user continues to type and the focus remains in the typing area. Actual behavior: After a few words, the focus shifts to the 'Tags' section and it continues to write in that section.
UX Framework PRB2054946	Keyboard shortcut remapping for admins and users	This is a product update.
Virtual Agent PRB1980327	A chat dynamic greeting isn't localized correctly	The message doesn't go down the correct API path to resolve to sys_ui_message, so it can't honor the dynamic greeting.
Virtual Agent PRB2009200	For the 'Resume' flow, work notes are added as the guest user		Trigger the ZTSD flow with a query that has a valid KB resolution available. Impersonate as the subject person. Ask a query that can be answered by an existing KB article. The 'Resume' flow is triggered. Open the corresponding case and check the 'Work Notes' section. Expected behavior: Work notes should be added under the impersonated HR L1 worker. Actual behavior: Work notes are added under the guest user.
Virtual Agent PRB2026682	Conversation history of language detection confirmation disrupts a user response	According to Search QnA, the prompt isn't structured to support the intermediate language selection turn. That should be removed from the conversation history.	Configure language detection. Start a conversation in any standard chat. Type an utterance in a different language. When presented with 'You are typing in xyz language, do you want to switch', select Yes. Expected behavior: The language should switch and the synthesized response should be displayed. Actual behavior: The language switches but the response is 'How can I help you with your laptop'.
Virtual Agent PRB2031232	The 'Litjs' widget isn't displaying on the Now Assist panel
Virtual Agent PRB2031583	Non-topic skills are dropped from the skill picker when all applicable topic skills have a visible design category	The picker shows the topic skills grouped under their design categories, but the non-topic skills are absent. No 'Others' category is created and the non-topic skills aren't rendered.	Configure a VA/NowAssist conversation with: At least one non-TOPIC skill that is applicable and visible (optionally, mark it as promoted in the context profile). At least one TOPIC-type skill that is applicable and visible. Ensure every applicable topic skill has at least one visible design category (sys_cb_topic_category with visible = true). Trigger the skill picker (for example, via SystemScriptObject.jsFunction _sendSkillPickerControl → VASkill​Service.​send​Skill​Picker)​.​ Expected behavior: Non-topic skills appear in the picker regardless of whether the applicable topic skills all carry visible design categories. For example, they could appear under 'Others'. Actual behavior: Non-topic skills are omitted whenever there is more than one applicable topic skill and all applicable topic skills have a visible design category.
Virtual Agent PRB2031991	'Get details of problem agent' returns additional duplicate closure messages along with the expected closing message	When the user enters a closing utterance, the agent returns multiple closure messages, and the last two messages are duplicated.	Open NAP. Enter an utterance like 'Help me get details of a problem'. Observe that the agent returns a response like 'Could you please provide the problem number you'd like details for?'. Enter an utterance like 'See you later'. Expected behavior: The agent returns a response like 'Understood. Feel free to return anytime if you need help.' Actual behavior: Additional closure messages show up, and the last two messages are duplicated. For example, the agent might return the following: 'Understood. Feel free to return anytime if you need help. Thanks for chatting! I'll go ahead and close this conversation now. I'm here if you need anything else. It looks like you're finished with this chat, so I'll go ahead and close it. It looks like you're finished with this chat, so I'll go ahead and close it.'.
Virtual Agent PRB2035888	A Guardian-triggered async_search early-return leaves a stale task ID in the context		Send a prompt that Guardian flags. Observe that it's displayed correctly. Also, observe that the first request is still active but times out later. Send a second utterance that triggers skill Discovery immediately. Wait for a timeout from the first message to occur to display a 'sorry' message in the conversation. Expected behavior: The second utterance responds normally. Actual behavior: 'Sorry, there was a problem on my side' appears. The second utterance result is dropped.
Virtual Agent PRB2037636 KB3108272	Agent messages containing URLs aren't displaying correctly in the internal transcript		Refer to the listed KB article for details.
Virtual Agent PRB2054519	Rename five assistant names from 'Now Assist' to 'Otto'	The old assistant names should be renamed as following: 'Now Assist in Virtual Agent' > 'ServiceNow Otto for Virtual Agent', 'Now Assist Panel – Platform' > 'ServiceNow Otto panel – Platform', 'Now Assist Panel – Developer' > 'ServiceNow Otto panel – Developer', 'Now Assist Voice Deployment' > 'ServiceNow Otto voice', 'Now Assist in Virtual Agent - N' > 'New chat assistant - 1', and 'Now Assist Voice Development - N' > 'New voice assistant - N'.
Virtual Agent PRB2055706	Snc_internal role is missing on sys_cs_conversation write ACL when the 'Explicit Roles' plugin is also active	Since the snc_internal role is never associated to the ACL, any users with snc_internal role can't upload files from NextWave conversation experience. The /api/now/upload/attachment API checks for write access on the actual record itself before allowing file upload, so the upload API is failing.	Provision a Zurich instance with the 'Explicit Roles' plugin (com.glide.explicit_roles) and the Conversation Server plugin (com.glide.cs) installed. Navigate to an ACL on sys_cs_conversation table with write operation on record. Observe that only the snc_external role is added. The snc_internal role is never associated to this ACL.
Virtual Agent PRB2056495	Central cache doesn't return the correct entry if the cache is updated from a different cluster	In the central cache log, the following message appears: 'Negative cache hit — Glide previously returned null for this key'.
Virtual Agent PRB2057596	Rename 'Now Assist Virtual Agent' to Otto for topics in Glide
Virtual Agent PRB2058703	Generating a KB article throws an error	The following error appears: 'Configured callback URL for the KB generation topic is invalid: https:​/​/​nextwave-​preview-​internal- ​c003.​aus100.​service-​now.​com'.​	Create an instance, following the documentation for setting up NextWave instances. Navigate to now-assist-admin. Make sure the KB Generation skill is active and the display for NAP(OTTO) channel is active. Give the prompt 'Create the KB for INCXXXXXXX' or 'Generate KB article'. Observe that there are three potential outcomes. First, the AI output could be: 'I wasn't able to generate the KB article because the configured callback URL for the KB generation topic is invalid'. Second, the AI output could be: 'I'll help you create a knowledge article. Let me first retrieve the incident details to understand what information should be included in the KB'. Then the execution stops. It doesn't collect any information nor create any article. The error code is 200102. Finally, it might not use the KB Generation skill. Instead, it uses knowledge graph or basic reasoning to form a draft article in the NAP window.
Virtual Agent PRB2058728	Chat summary isn't generated on instances with the Spanish translation plugin	After ending a chat between a requester and an agent, the summary doesn't generate. This occurs when one person is using Spanish and the other is using English.	Provision an instance with the Spanish translation plugin installed. Set Spanish as the language for the requester. Set English as the language for the agent. Make sure DT is enabled. As the requester, initiate a chat and connect with the agent. Exchange around 20 messages. As the requester, end the chat. Observe that the chat summary doesn't generate.
Virtual Agent PRB2059089	The Otto session isn't aware of the logged-in user for 'assets managed by me' queries	The Otto/AICT chat session isn't aware of the logged-in user by default. When a user asks a question like 'Can you give me the assets managed by me?', the assistant returns the complete/unfiltered list of assets instead of applying a managed_by = current user filter. The filter is only applied when the user explicitly names themselves in the query. The assistant should recognize 'me'/'my' references and automatically apply the logged-in session user's context without requiring explicit clarification.	Log in as an AICT user. Ask Otto: 'Can you give me the assets managed by me?'. Observe that the full/unfiltered asset list is returned instead of being filtered to the current user. Ask explicitly by name, for example: 'assets managed by <username>'. Observe that the managed_by = current user filter is correctly applied.
Virtual Agent PRB2059632	BuildPolicyConfig should be aligned with sys_​now_​assist _​va_​persona_​detail schema	No policy configs are returned because buildPolicyConfig doesn't target sys_​now_​assist _​va_​persona_​detail and doesn't filter by persona_detail_type = Policy.	Create an active record in sys_​now_​assist_​va_​persona_​detail:​ Deployment = a valid deployment sys_id. Persona_detail_type = Policy. Name = Test Policy. Description = Test policy description. Prompt_value = Test policy instruction. For the same deployment, create another active persona-detail record with persona_detail_type = Tone and prompt_value = Friendly. Trigger the assistant-config handshake for that deployment via next wave. Inspect policyConfigs in the response. Expected behavior: One policy config entry is returned, containing the sys_id, name, description, active, and instruction from the policy persona-detail record. Non-policy persona-detail records are excluded. Actual behavior: No policy configs are returned because buildPolicyConfig doesn't target sys_​now_​assist_​va_​persona_​detail and doesn't filter by persona_detail_type = Policy.
Virtual Agent PRB2060897	The 'promoted-skills' API doesn't return non-discoverable skills		Create a skill. Call the 'promoted-skills' API. Expected behavior: The skill should be returned. Actual behavior: The skill isn't returned.
Virtual Agent PRB2061042	Fixes are needed for guest user support for OffGlide	Additional rest endpoints need a public role. The NextWave token needs to be turned on for bff exchange.
Virtual Agent PRB2061520	Calling the 'Bg channel' API with the same session ID isn't resuming the conversation		Create a incident and assign it to the ZTSD worker. Once solution is proposed, reply with new comments. Expected behavior: The conversation, which is waiting for input, should resume back. Actual behavior: Creating a conversation and execution plan with the workflow as 'Default VA Workflow'.
Virtual Agent PRB2063396	Parallel tools execution aren't running in record domain
Virtual Agent PRB2064355	Auto Chat executions are faulted	There is an intermittent failure in fetching the JWT token from CS, which is causing the faulted conversations.
Virtual Agent PRB2064833	A domain is set to null after user input
Virtual Agent PRB2066077	Add the reduce_items_list_polling toggle
Window Manager PRB2052078	Resize doesn't work at run time, only on creation time	The talk modal needs to be a fixed size. The user dragging it smaller breaks the layout, so the grip should be disabled at runtime.	Try {canResize: false}. Expected behavior: The gripper disappears. Actual behavior: There's no change.
Window Manager PRB2052105	SetWindowProperties ignores empty headingLabel and it can't clear a window's title at runtime	The previous title stays. '' and null are silent no-ops.	Open a window with a heading (for example, the canvas window with any headingLabel). Call setWindowProperties(windowId, { headingLabel: '' }). Expected behavior: The heading clears, resulting in a blank title. Actual behavior: The previous title stays. '' and null are silent no-ops.
Work Order Management PRB2035667	Intraday optimization is stuck in 'In progress'	During normal intraday runs with concurrent solution processing, the race condition between SolutionMetadataProcessor and SolutionProcessorSNC can cause lost updates on ml_solution.solution_metadata. This results in stuck in_progress records.	Configure Schedule Optimization with intraday optimization enabled with territory model. Set up an intraday optimization schedule with multiple overlapping territories. Trigger an intraday optimization run. Check wm_intraday_job_m2m_qualifier. Observe that some of the records remain stuck in the in_progress status indefinitely, even after all the solutions are processed or the records show no processed task counts, total task counts, and/or progress bar.

Fixes included

Unless any exceptions are noted, you can safely upgrade to this release version from any of the versions listed below. These prior versions contain PRB fixes that are also included with this release. Be sure to upgrade to the latest listed patch that includes all of the PRB fixes you are interested in.

• Zurich patch 11 Hotfix 2
• Zurich Patch 11 Hotfix 1
• Zurich Patch 11
• Zurich Patch 10 Hotfix 5
• Zurich Patch 10 Hotfix 4a
• Zurich Patch 10 Hotfix 4
• Zurich Patch 10 Hotfix 3a
• Zurich Patch 10 Hotfix 3
• Zurich Patch 10
• Zurich Patch 9 Hotfix 6
• Zurich Patch 9 Hotfix 5
• Zurich Patch 9
• Zurich Patch 8 Hotfix 5
• Zurich Patch 8
• Zurich Patch 7b Hotfix 3
• Zurich Patch 7
• Zurich Patch 6
• Zurich Patch 5
• Zurich Patch 4
• Zurich Patch 3
• Zurich Patch 2
• Zurich Patch 1
• Zurich security and notable fixes
• All other Zurich fixes