Consolidated page of all release notes for Generative AI Controller from Washington DC to Zurich.
How to use this page
To help you prepare for your upgrade, we have combined the cross-family Generative AI Controller release notes onto one page. Read this summary of the new features, changes, and updated information for your product from Washington DC to Zurich.
Tip: If there were no updates for a release notes section in a certain family release, we included a short note for your reference. For example, if a product did not have any updates in Tokyo, the row says "No updates for this release."
Important information for upgrading Generative AI Controller to Zurich
Before you upgrade to Zurich, review these pre- and post-upgrade tasks and complete the tasks as needed.
| Release |
Release notes |
Washington DC |
No updates for this release. |
Xanadu |
No updates for this release. |
Yokohama |
Generative AI Controller is installed and updated when you install or update any Now Assist application. If you have issues installing or updating applications, see this knowledge article for steps that may address your issue. Otherwise, you can make a Support case.
|
Zurich |
Generative AI Controller is installed or updated when you install or update a Now Assist application. If you have issues installing or updating applications, see this knowledge article or open a support case.
|
New features
Between your current release family and Zurich, new features were introduced for Generative AI Controller.
| Release |
Release notes |
Washington DC |
- Recursive summarization for large inputs
- Large language models (LLMs) have a maximum number of tokens per request. Recursive summarization breaks down large inputs into smaller pieces, summarizes those pieces, and then recombines them to make the original call.
This process prevents important context from being lost when submitting requests to the LLM, which can result in more accurate responses with fewer hallucinations.
- Integrate with Aleph Alpha models
- Bring your own API credentials and connect Aleph Alpha's Luminous LLM with the ServiceNow AI Platform.
- Integrate with IBM watsonx models
- Use your own API credentials to connect to IBM watsonx's Granite LLMs with the ServiceNow AI Platform.
- Integrate with external LLMs
- Connect API credentials with a generic LLM connector and use an external LLM on the ServiceNow AI Platform.
- Semantic filter for sensitive content in generative AI conversations
- Use AI Semantic Search to help determine if sensitive content is being discussed, such as workplace misconduct or employee health concerns, and trigger a switch to a different Virtual Agent topic.
- Disable Dynamic Translation for LLM conversations in Virtual Agent
- Dynamic Translation is enabled by default for ServiceNow AI Platform feature that is available with activation of any Now Assist applications, but you can disable it if you prefer not to have multi-language support for LLM conversations.
|
Xanadu |
- Xanadu Patch 3
- Metadata log table for generative AI requests
- Use the new metadata log table that includes information about the requesting conversation, the capability definition, provided feedback (if any), and error codes.
- Translation for Now Assist
- Support users who speak different languages with new language information in LLM request metadata, better controls for Dynamic Translation, and translated text logging.
- Xanadu Patch 7
- Global Model selection for conversational skills
- Enable Now Assist Admins to choose between GPT4.o and Now Assist LLM model for data routing at a global level. This ensures compliance with any regional restrictions and helpful for APAC users who may face limitations with US-based models, such as GPT-4.0.
|
Yokohama |
- yokohama-patch-13
- [Placeholder link text to key configure-custom-resource-path-byok]
- Connect Generative AI Controller to your Azure OpenAI deployment by configuring a custom resource path in your bring your own key (BYOK) model configuration. Use this when your Azure OpenAI endpoint includes a path segment that Generative AI Controller does not add by default.
- Yokohama Patch 10
- Administrator access to Gen AI log
- Access the Gen AI log [sys_generative_ai_log] table to gain insights for debugging purposes. Records related to HR remain restricted to HR admins.
- Enhanced AI asset inventory
- Track the enhanced AI asset inventory through AI Control Tower using new metadata fields in the Model [sys_generative_ai_model_config] and Prompt [sys_generative_ai_config] tables. Gain better visibility into AI asset status and life-cycle details, such
as retirement dates.
- Yokohama Patch 8
- AI Model Version Mappings
- Review the mappings between AI model versions, their providers in the Gen AI Model Version Mapping [sys_gen_ai_model_version_mapping] table. It shows mapping between source and target models, along with associated metadata,
such as skill type, model type, resource associations, and provider information.
- Yokohama Patch 6
- Identify third-party LLM information
- Access the Gen AI Log Metadata [sys_gen_ai_log_metadata] table to identify which LLM model, version, and requested language was used to generate the AI content.
- Restrict LLM usage based on the domain
- Enable or disable Now Assist for each domain so that you can restrict the use of LLMs and avoid using AI for data processing, if needed.
- Yokohama Patch 1
- Global Model selection for conversational skills
- Enable Now Assist Admins to choose between GPT4.o and Now Assist LLM model for data routing at a global level. This ensures compliance with any regional restrictions and helpful for APAC users who may face limitations with US-based models, such as GPT-4.0.
|
Zurich |
- Zurich Patch 8
- Configure a custom resource path for BYOK models
- Connect Generative AI Controller to your Azure OpenAI deployment by configuring a custom resource path in your bring your own key (BYOK) model configuration. Use this when your Azure OpenAI endpoint includes a path segment that Generative AI Controller does not add by default.
- Zurich Patch 4
- Administrator access to Gen AI
log
- Access the Gen AI log [sys_generative_ai_log] table to gain insights for debugging purposes. HR-related records remain restricted to HR admins.
- Enhanced AI asset inventory
- Track the enhanced AI asset inventory through AI Control Tower using new metadata fields in the Model [sys_generative_ai_model_config] and Prompt [sys_generative_ai_config] tables.
Gain better visibility into AI asset
status and life-cycle details, such as retirement
dates.
- Long-Term Stable model for generative AI
- Promote stability and adoption for entitled users with a Long-Term Stable (LTS) model. The LTS model provides regulatory alignment, predictable behavior, and life
cycle stability to integrate GenAI with confidence.
- Zurich Patch 2
- AI Model Version Mappings
- Review the mappings between AI model versions and their providers in the Gen AI Model Version Mapping [sys_gen_ai_model_version_mapping] table.
This table shows the mappings between source and target models, along with associated metadata, such as skill type, model type, resource associations, and provider information.
- Identify third-party LLM information
- Access the Gen AI Log Metadata [sys_gen_ai_log_metadata] table to identify which LLM model, version, and requested language was used to generate the AI content.
- Restrict LLM usage based on the domain
- Enable or disable Now Assist for each domain so that you can restrict the use of LLMs and avoid using AI for data processing, if needed.
|
Changes
Between your current release family and Zurich, some changes were made to existing Generative AI Controller features.
| Release |
Release notes |
Washington DC |
No updates for this release. |
Xanadu |
|
Yokohama |
|
Zurich |
|
Removed
Between your current release family and Zurich, some Generative AI Controller features or functionality were removed.
| Release |
Release notes |
Washington DC |
- The QnA capability has been removed and the action has been removed from Virtual Agent Designer and Workflow Studio.
|
Xanadu |
No updates for this release. |
Yokohama |
No updates for this release. |
Zurich |
No updates for this release. |
Deprecations
Between your current release family and Zurich, some Generative AI Controller features or functionality were deprecated.
| Release |
Release notes |
Washington DC |
No updates for this release. |
Xanadu |
No updates for this release. |
Yokohama |
The dependency on Sensitive Data Handler has been removed. Regular expressions that are configured with Sensitive Data Handler are applied to the Data Privacy application with a fix script when you upgrade to Yokohama.
|
Zurich |
No updates for this release. |
Activation information
Review information on how to activate Generative AI Controller.
Additional requirements
If any additional requirements were introduced or changed for Generative AI Controller we have noted them here.
| Release |
Release notes |
Washington DC |
No updates for this release. |
Xanadu |
No updates for this release. |
Yokohama |
No updates for this release. |
Zurich |
No updates for this release. |
Browser requirements
If any specific browser requirements were introduced or changed for Generative AI Controller we have noted them here.
| Release |
Release notes |
Washington DC |
No updates for this release. |
Xanadu |
No updates for this release. |
Yokohama |
No updates for this release. |
Zurich |
No updates for this release. |
Accessibility information
Review details on accessibility information for Generative AI Controller, such as specific requirements or compliance levels.
| Release |
Release notes |
Washington DC |
No updates for this release. |
Xanadu |
No updates for this release. |
Yokohama |
No updates for this release. |
Zurich |
No updates for this release. |
Localization information
If there are specific localization considerations for Generative AI Controller we have noted them here.
| Release |
Release notes |
Washington DC |
No updates for this release. |
Xanadu |
Generative AI Controller uses Microsoft Azure OEM for Dynamic Translation in Now Assist for multi-language support. You can enable dynamic translation from the Now Assist Admin console.
|
Yokohama |
No updates for this release. |
Zurich |
Generative AI Controller uses Microsoft Azure OEM for Dynamic Translation in Now Assist for multilanguage support. You can enable dynamic translation from the Now Assist Admin console. For more information, see Microsoft Azure OEM for Dynamic Translation in Now Assist.
|
Highlight information
If there are specific highlight considerations for Generative AI Controller we have noted them here.
| Release |
Release notes |
Washington DC |
- Integrate with custom LLMs in generative AI capabilities in Workflow Studio and Virtual Agent Designer
- Add semantic filters in Virtual Agent to detect and handle sensitive topics
See Generative AI Controller for more information.
|
Xanadu |
Xanadu Patch 3
- Review a read-only metadata log table for generative AI requests.
- Use the improved language support for skills and capabilities, including the Dynamic Translation services and native translation by the LLMs.
Xanadu Patch 1
- Updated prompts for capabilities, semantic filtering, and recursive summarization that improve latency and processing for Generative AI Controller capabilities and Now Assist skills.
See Generative AI Controller for more information.
|
Yokohama |
Yokohama Patch 11
- Review changes to Now Assist usage measurement.
- Some Now Assist skills, agents, and agentic workflows are on by default.
- Additional role configuration is required for agentic workflows and AI agents included with Now Assist applications.
Previous Patch releases
- Protect your users by configuring settings for anonymization of personally identifiable information (PII) with the Data Privacy application.
See
Generative AI Controller for more information.
|
Zurich |
Zurich Patch 8
- Connect your Azure OpenAI deployment to Generative AI Controller by configuring a custom resource path in your bring your own key (BYOK) model configuration.
Zurich Patch 4
- Access the Gen AI log for debugging insights, with HR-related records remaining restricted.
- Promote stability with the Long-Term Stable (LTS) model for generative AI.
- Identify third-party LLM information, including model, version, and language.
- Restrict LLM usage based on domain.
See Generative AI Controller for more information.
|