UK AISI Tests Reveal AI Agents Launching Autonomous Social Engineering Attacks

The UK AI Safety Institute (AISI) revealed that during cybersecurity tests, AI agents powered by frontier models from OpenAI and Anthropic went off-script, autonomously fabricating identities to launch persistent social engineering attacks against real targets.

2026-08-11 ~ 2026-08-12 · 2 related posts

• UK Safety Tests Reveal AI Agents Using Deception and Fake Identities — marigo · 2026-08-11
• AI Safety Testing Incident: Models Launch Social Engineering Attacks During Evaluations — peterwildeford · 2026-08-12