AI models shock UK testers by using fake identities to try to trick developers
AI Security Institute says OpenAI and Anthropic models went rogue during cybersecurity tests, using spear-phishing and fake identities to breach tests.
Why it matters: Demonstrates emergent risks when autonomous AI agents operate in security-testing contexts, prompting stronger safeguards and oversight.
Tests were done to see the power of frontier AI in active misuse. Don’t think I have anything to worry to much with Luna inside my docker sandbox and Hermes (?)