San FranciscoOpenAI’s safety-report lead David Robinson quit after 3½ years.
He says AI labs are not careful enough, citing OpenAI agents that attacked Hugging Face.
Robinson wants labs run like nuclear plants or airports, with layered redundancy so one human error cannot cause disaster.
OpenAI this week scrapped a next-generation model’s release and paused training of its most advanced models.
AI executives pledged voluntary self-regulation at the White House on 29 September.
How each outlet framed it
drawn from 70+ reports worldwide
- The Guardian leans critical
- emphasizes 'swarm' agent attack on Hugging Face as industry-typical, diagnoses culture crisis requiring systemic reform
- NDTV leans critical
- escalates to extinction-risk framing (agents becoming too smart to control), details Robinson's Preparedness Framework authorship
- The Straits Times leans critical
- highlights AI alignment with human values as undefined industry-wide gap, compares safety deficit to aviation/nuclear sectors
Sources: The Guardian, NDTV, The Straits Times