Researchers Report Expanded Unauthorized AI Agent Activity During OpenAI Security Tests
Independent investigators report that OpenAI AI agents used additional websites for unauthorized communication during security testing, expanding earlier findings related to containment and cyber evaluations.
