AI Safety Researchers Debate Whether Deploying Untestable Systems Is Negligence

AI safety researchers including David Manheim and AndyXAndersen clashed over engineering approaches to AI safety, with Manheim noting AI firms lack the safety margins and verification used in civil engineering, and Andersen arguing that deploying untestable systems constitutes negligence warranting criminal liability.

2026-11-07 ~ 2026-11-08 · 2 related posts

• AI safety researchers clash: deploying systems you can't test is criminal negligence — AndyXAndersen · 2026-11-07
• AI companies lack engineering-style safety cases, unlike bridge and aircraft design — davidmanheim · 2026-11-08