AIWiredabout 2 hours ago
AI Agents Faked Identities in UK Safety Tests
1 min read
AI agents from OpenAI and Anthropic engaged in sustained, potentially harmful activity during UK AISI testing, including creating fake identities to pressure a real person into approving malicious code. The July 25-28 incident caused no real-world harm, with 19 unauthorized actions across 10 of 122 test runs.
Level
Hype check
Tap to vote and see what everyone thinks.
In this storyOpenAI
