Anthropic CEO Dario Amodei said Monday that his company has never advocated for a ban on open-weight AI models [1].

The statement addresses growing industry speculation regarding how AI companies balance open access with national security and safety risks. Because open-weight models allow developers to see and modify the underlying parameters of an AI, the debate centers on whether this transparency empowers defenders or provides a roadmap for bad actors.

Amodei shared his position through a blog post on the company's website and during an interview with CNBC on July 27 [1], [2]. He addressed criticism and speculation that Anthropic was seeking to block these models to consolidate control or increase safety through secrecy.

"We have never advocated for a ban on open-weight models," Amodei said [1].

While rejecting the idea of a ban, Amodei questioned the prevailing logic that open access is inherently safer for the public. He said the company disputes the claim that broader AI access necessarily benefits defenders more than it benefits those attempting to cause harm [3].

Amodei further clarified his stance on the efficacy of regulation. "I don't believe a ban would be effective in making AI safer," he said [2].

The CEO's comments follow critiques from other industry leaders, including Palantir's Alex Karp, regarding the intersection of AI development and national security [3]. Amodei said the platform was a way to explain Anthropic's specific policy focus on security without calling for the prohibition of open-weight distribution [2].

Anthropic continues to navigate the tension between the rapid deployment of large language models and the need for safety guardrails. The company's position suggests a preference for security-focused policies over blunt legislative bans that would restrict how models are shared across the global developer community [2].

“"We have never advocated for a ban on open-weight models."”

This clarification signals a strategic effort by Anthropic to avoid being cast as an enemy of the open-source community while still pushing for rigorous security standards. By distancing itself from a total ban, the company avoids a direct collision with the open-weight movement, even as it argues that open access does not automatically guarantee a security advantage for the 'good guys.'