Will Brown: regulating AI models by 'capability' can't even draw the line
willcb · x · 2026-10-03
Will Brown pushes back on conflating his argument with copyright DMCA: he's talking about regulating model weights on the basis of being a "powerful AI tool that could potentially be used for some antisocial thing".
His core points:
- Gambling and drug precursors have fairly consensus definitions, yet regulating gambling is still hard at the margin since the line between gambling and investing is blurry;
- By analogy, what separates an AI girlfriend, friend, therapist, or coworker? Should we ban Claude? All voice models?
- What has worked best historically is regulating at the marketing level, which he doesn't oppose.
Related event: willcb on AI weight regulation: meaningful legislation won't arrive in time(2 posts)→
More from Safety
- Personal agents' broad macOS access faces an Apple crackdown, warns dev — signulll · 2026-10-03
- NetScaler SAML exploit exploited in the wild; open-source checker scans 8 CVEs and 93 attacker IPs — cyb3rops · 2026-10-03
- Stanford HAI's James Landay on educating lawmakers and shaping AI with people at the center — StanfordHAI · 2026-10-03
- CAIS founder Dan Hendrycks launches AI safety intro course on Coursera — irinarish · 2026-10-03
- EU's homegrown Teams rival Element Pro panned by its own officials as 'absolute shit' — castrotech · 2026-10-03
- Flock Safety threatens legal action against developer behind FlockSurveillance.org — Polymarket · 2026-10-03