The United States has removed a licensing requirement that barred Anthropic from exporting its flagship AI models, Mythos and Fable, to foreign users. The restriction, imposed on June 12, classified the models as export‑controlled technology, effectively ending public access and compelling Anthropic to pull the models from its platform.
After weeks of negotiations, Secretary of Commerce Howard Lutnick announced that Anthropic "has agreed to proactively detect and address security risks associated with the models; to work diligently with the U.S. government on protocols and standards and releases for Mythos, Fable and future models; and to inform the U.S. government of any malicious activity." The company had already pledged to adopt many of those safeguards voluntarily before the rule took effect.
Commerce Department clears models for vetted customers
Under the new arrangement, Anthropic will begin restoring access to Mythos on July 1, starting with a limited set of customers cleared by the White House. The move mirrors a similar clearance granted to OpenAI’s latest models, which are also being distributed to a select group of organizations rather than the general public.
Anthropic’s Mythos was first released to a handful of organizations in April as a test to allay concerns about its ability to discover software vulnerabilities. Fable followed in June, rolled out to the public with additional security guardrails. Critics of the export ban argued that it served more as political leverage than a genuine security measure, especially after Anthropic’s executives publicly criticized how the government and the president’s opponents might exploit the technology.
Pressure to ease the restrictions grew as Asian AI firms, including Fugu and Tulongfeng, announced models that approach Mythos‑level capabilities. U.S. officials faced mounting calls to ensure American AI remains competitive on the global stage.
The administration’s approach to AI policy has been described as erratic, leaving companies uncertain about future release guidelines. A June executive order that called for pre‑release model reviews drew criticism from analysts such as Dean W. Ball, who recently joined OpenAI’s policy team.
With the licensing hurdle removed, Anthropic can once again serve international clients, albeit under a framework that emphasizes risk monitoring and government coordination. The decision signals a shift toward a more collaborative, though still cautious, stance on AI export controls.
Cet article a été rédigé avec l'assistance de l'IA.
News Factory APP - actualités agentiques pour booster votre SEO et AEO.