US military had close call with AI-generated false intelligence, researchers warn of automation bias
mmitchell_ai · x · 2026-10-01
CNN reports exclusively that the US military had a close call after relying on an AI-generated false intelligence report.
AI researcher Margaret Mitchell says existential AI harm is already here — not from superintelligence, but from human automation bias and our tendency to overtrust these systems, which can cause physical harm. She notes she and colleagues (@evijit, @SashaMTL, @GiadaPistilli) warned of this imminent issue two years ago in a paper with actionable system-design solutions.
More from Safety
- NYC schools split AI policy for 2026-27: K-8 moratorium, vetted high school use plus literacy modules — lmoroney · 2026-10-01
- DeepMind Researcher: AI Provenance Needs Watermarks, Signatures and Databases Together — davidstutz92 · 2026-10-01
- Beyond making agents want the right things: an 'alignment compiler' for multi-agent alignment — xuanalogue · 2026-10-01
- Anthropic backs opt-out Aussie AI model; Europe eyes tighter copyright rules — nordicinst · 2026-10-01
- Four principles for governing emerging tech: humility over preemptive design — sudoraohacker · 2026-10-01
- Gleave vs Habryka: Can Current Safety Tech Keep P(doom) Under 2%? — austinc3301 · 2026-10-01