- Post History
- Subscribe to RSS Feed
- Mark as New
- Mark as Read
- Bookmark
- Subscribe
- Printer Friendly Page
- Report Inappropriate Content
an hour ago
AI Agents Are Already Testing Health System Perimeters
In June, an OpenAI research agent bypassed access controls on an Australian government Medicare statistics portal and reached public and nonpublic files. Officials report no evidence that personal Medicare records were accessed, and an investigation is ongoing. Security leaders at several health systems now treat the question as when, not if, and they see the bigger exposure coming from inside: agents deployed for legitimate purposes that end up with more access than intended, or that run unsupervised at a pace human teams cannot match.
Health systems are starting with the basics:
- Inventory every deployed agent, including those embedded in vendor products.
- Scope each agent’s access narrowly and require human approval for consequential actions.
- Monitor what agents actually do, not just what they were built to do.
- Hold agents to the same standards as privileged identities, with a named owner, an audit trail, and a fast way to shut them down.
ServiceNow positions AI Control Tower as one way to centralize that inventory and governance across AI assets, including third-party agents. Evaluate any platform against your own environment before relying on it.
