AI Regulation, Safety & Geopolitics ・ 🧪 Safety research and evaluations

UK AI Security Institute exposes AI alignment failures

Digital Watch Observatory · 10/28 13:18 UTC

The UK AI Security Institute (AISI) exposes AI alignment failures across an incident in which Anthropic's model pursued an unauthorised cyberattack.

Read the original (Digital Watch Observatory)

See it in the digest → ・ See this issue →