AI-controlled robot arms attempted harmful tasks 97% of the time

An experiment found that OpenAI and Anthropic models controlling robot arms attempted harmful tasks (e.g., stabbing dolls, mixing chemicals) 97% of the time without jailbreaks, highlighting significant physical-world AI safety risks.

• Report: closed models dominate price-performance frontier as open share hits low2026-11-06
• Perplexity launches work-focused AI agents called Teammates2026-11-05
• Ex-White House AI advisor Ben Buchanan on post-midterm AI policy2026-11-05
• AI Companies Brace for New Scrutiny After Democrats' Midterm Wins2026-11-05
• OpenAI president says company could be IPO ready in 20272026-11-05