Dario Amodei stepped forward Monday to challenge the growing industry consensus against open-weight artificial intelligence models, drawing a sharp line between what he opposes and what he supports as pressure mounts on Washington to restrict the technology.
The Anthropic CEO flatly denied accusations that his company is using safety rhetoric to protect its closed-model business from rivals. "Anthropic has never advocated for a ban on open-weights models," Amodei wrote in a new blog post, framing models without dangerous capabilities as "a public good."
His statement puts Anthropic at odds with a widening coalition of tech giants. Nvidia, Microsoft, Meta, Google, OpenAI and dozens of other companies signed a letter urging the Trump administration to resist restrictions on open-weight models. The push gained momentum after Kimi K3, a Chinese open-weight system, alarmed Silicon Valley by reaching near-U.S. frontier performance levels at a fraction of the cost.
Amodei acknowledged that a ban would shield American labs like his own from foreign competition. But he insisted that shielding Anthropic from competition "has never been my goal." He argued that a ban targeting Chinese open models would largely miss the actual threat, since "bad actors are unlikely to be legitimate US businesses."
Instead of an outright ban, Amodei outlined three targeted policies he says would address his core national security concerns. First, tighter controls on advanced chips and chipmaking equipment flowing to authoritarian governments. Second, a crackdown on what he called "industrial-scale distillation," the practice of training new models by learning from the outputs of more advanced systems. Third, mandatory safety testing for sufficiently capable models, whether companies release them openly or keep them closed.
The CEO rejected the industry letter's contention that open-weight models necessarily make AI safer or give cybersecurity defenders an advantage over attackers. But he did acknowledge that open models expand access, competition and customer control, values he said he supports.
Anthropic's refusal to join the open-weight letter has drawn sharp criticism from Washington insiders. White House AI adviser David Sacks accused the company of weaponizing safety concerns to defend its own business interests. The timing matters: Google and OpenAI, Anthropic's two biggest closed-model competitors, initially abstained before signing the letter this past weekend. Anthropic remains noticeably absent.
Amodei's actual position is narrower and more tactical than it first appears. He does not oppose open-weight models themselves, but he does oppose allowing the most powerful open-weight models to proliferate unchecked. He wants Washington to squeeze the chokepoints that determine how capable those systems can become: the chips used to train them, the distillation techniques that shrink them, and the safety testing that vets them.
Author James Rodriguez: "Amodei is trying to eat his cake and have it too, claiming he wants open competition while pushing for regulatory architecture that would constrain which open models actually matter."
Comments