BREAKING
Hassabis Pitches U.S.-Led AI Watchdog
Modeled on FINRA
FINRA
Model
●
Polices securities industry
●
Overseen by the SEC
AI Watchdog
Proposed
●
Industry-funded
●
Accountable to U.S. government
How Pre-Release Testing Would Work
1
Clear frontier benchmarks
↓
2
Share model up to 30 days
↓
3
Test catastrophic risks
↓
4
Voluntary to mandatory
Risks the Body Would Screen
1
Cybersecurity
↓
2
Biological threats
↓
3
Nuclear capabilities
↓
4
Signs of deception
Support Meets Skepticism
Support
●
Ties to other lab leaders
●
Path to international rules
Skeptics
●
Risk of regulatory capture
●
Reach over China unclear
Live Before End of 2026
AI NEWS BLITZ
Google DeepMind's Demis Hassabis wants a new body to police the world's most powerful AI.