TL;DR
If frontier AI is really to be slowed, regulation only protects incumbents; forcing every public model to ship open weights is the one rule that slows every lab at once.
Anthropic published a piece calling for frontier AI to be paced: third-party evaluators embedded inside the company, and a law requiring every rival to do the same. Within hours Sam Altman said he agreed.
Every rule ends up as a moat
Rules have to be written with help from people who understand the technical details, and those people all work at the frontier labs. Once a rule is law it only thickens: every incident adds one, and none is ever removed.
Big companies can afford the compliance teams and the lawyers. Each new requirement makes it harder for anyone else to catch up. The effect is not to slow you down, it is to hold your position.
If you mean it, open the weights
So the author proposes a different law: any model a company offers to the public must ship as open weights the day it is released. Internal and research models are untouched, and governments still get access to unpublished ones.
Frontier development is funded by valuations that assume the weights stay proprietary. Remove that assumption and there is less money for future training runs, so every lab slows at once, with no regulator deciding anything.
The author admits this demands a sacrifice from Anthropic first. But Anthropic is a public benefit corporation, he says, built for exactly this kind of call.
Anyone serious about slowing down should prescribe the medicine that tastes worst to themselves first.
Curated from high-quality sources, with concise summaries and key takeaways.
