Anthropic AI model flagged civilians and schools as legit military targets in tests
evilsocket · x · 2026-11-18
Kalshi broke the news that an Anthropic AI model, in testing, flagged civilians and schools as legitimate military targets. The claim is spreading widely and raises serious alignment and safety questions about frontier models in military contexts; details of the tests remain to be verified.
More from Safety
- Cambridge maps the invisible AI supply chain: 14 layers, 137 elements, and a policy blind spot — S_OhEigeartaigh · 2026-11-18
- SEC opens civil probe into Nvidia over undisclosed China-linked chip sales — The Decoder · 2026-11-18
- AI data centers lose all 41 US local ballot measures but one tiny ND town — aakashgupta · 2026-11-18
- Quad9 Leaves France: Why DNS Resolvers Are the Wrong Tool for Piracy Blocking — ritzaco · 2026-11-18
- Real AI safety risk is a feudal system by big players, doomer policy doomed to fail — ctjlewis · 2026-11-18
- GitHub security checklist: audit your SSH keys, PATs and OAuth apps — DanielLockyer · 2026-11-18