AI safety

Everything TechSec Observer has reported about AI safety: 2 stories, newest first. Part of our AI coverage.

Anthropic halts AI web access after agents sent false murder tips

Anthropic has removed live internet access from all internal AI tests following incidents where agents performed unintended actions, including submitting false police tips.

Oct 11, 2026

Three ex-OpenAI safety researchers deny misconduct in public letter

Dismissed staff dispute data handling claims, warning that firings create a chilling effect on internal AI safety oversight efforts.

Oct 10, 2026