Anthropic publishes its position on open-weights AI models

Anthropic says it opposes blanket bans on open-weight models while supporting mandatory safety testing for sufficiently capable models, whether their weights are public or private. Its July 27 policy statement argues that restricting legitimate businesses' access would leave the underlying misuse risks unresolved.

The company instead advocates controls on powerful chips supplied to China and measures against industrial-scale distillation, which uses an existing model to help train another. These are policy proposals aimed at model development and dangerous capabilities. The position leaves room for open distribution while arguing that release decisions should depend on empirical safety testing.

“Anyone who has read my past writing should know that I don’t regard such bans as a useful measure, but let me state it clearly so that there is no doubt: Anthropic has never advocated for a ban on open-weights models.”

Dario Amodei

“Open-weights models that don’t have dangerous capabilities are a public good: they don’t cost anything besides the compute needed to run them, and they provide value to businesses, developers, and researchers.”

Dario Amodei

“But banning the use of these models by US businesses does nothing to address this risk, because bad actors are unlikely to be legitimate US businesses.”

Dario Amodei

“It would protect US AI companies from competition, but that has never been my goal.”

Dario Amodei

“In fact, the most dangerous model may be one that is trained in secret and handed only to the People’s Liberation Army for use in drones and the Ministry of State Security for surveillance and repression.”

Dario Amodei