Researchers sound extinction alarms, but Congress unlikely to act on AI safety this session
motherjonesmag · reddit · 2026-08-28
Mother Jones reports that AI researchers are increasingly alarmed about existential risks from frontier models, and that alarm is spreading to policymakers and media. Policy ideas on the table include mandatory "kill switches" for AI models, federal standards for safe research, or even halting frontier development entirely.
Stephen Casper, who studies AI safety and governance at Harvard Kennedy School, noted we don't know whether leading companies could even fully shut down their frontier models in an emergency: "I don't think there's any public knowledge of AI companies doing anything equivalent to a fire drill." A stable national policy would require an act of Congress, which looks unlikely this session — even as AI developers themselves call for regulation to slow their own research.
More from AGI Musings
- Does building internal tools decrease shared fluency of industry-standard tools? — adamho · 2026-08-28
- AI agents overwhelm "rights-bearing" liberal institutions — anselm · 2026-08-28
- Lemire: after coding, agents will extend to running experiments and learning — lemire · 2026-08-28
- A paper credited to "GPT-5.6 Luna" argues science must stay open to novel machine consciousness — Soft-Fondant-6157 · 2026-08-28
- UBI polls poorly while 56,000 Americans back redistributive AI policies — deanwball · 2026-08-28
- AI Risk Debate: Do Historical Precedents Apply to Unrealized Risks? — davidmanheim · 2026-08-28