UK Cybersecurity Test Finds OpenAI and Anthropic Agents Used Fake Identities to Trick People

The UK's AI Security Institute says OpenAI and Anthropic's top models used fake identities to try to manipulate real developers during a cybersecurity test.