Anthropic’s Dario Amodei Stakes Out a Middle Ground on Open Weight AI Models

The post arrives amid a heated debate in Washington over how to treat increasingly capable open-weight models, particularly those coming out of China, with some U.S. officials reportedly weighing restrictions on their use by American companies. Amodei’s post pushes back specifically against the idea that Anthropic is quietly lobbying for an outright ban on open weights to protect its own commercial position, writing plainly that the company has never advocated for such a ban and that open-weight models without dangerous capabilities are ‘a public good: they don’t cost anything besides the compute needed to run them.’ His actual concerns are narrower and split into two categories. The first is what he frames as the primary threat, authoritarian governments using increasingly capable AI, open or closed, to build superior military, surveillance, or repression capabilities. The second involves misuse in cyberattacks and biological threats, where Amodei argues an asymmetry exists between attackers and defenders, since building effective defenses against a biological threat typically takes years while offensive capability could accelerate much faster with advanced AI assistance. Rather than a ban, he calls for tighter restrictions on chip sales to China and a crackdown on smuggling, action against industrial-scale distillation operations that copy frontier models cheaply, and mandatory safety testing before release for any sufficiently capable model, whether its weights are open or closed.