US Freezes Then Reopens Anthropic Models

Anthropic just faced the new AI reality: Washington abruptly shut down access to its most advanced AI models for two weeks, then turned it back on after a tense round of negotiations over security controls. At the heart is whether companies like Anthropic can keep selling their top-tier models—like Fable 5 and Mythos 5—when the U.S. government demands tighter safeguards to prevent misuse. This isn’t just a policy scuffle: the ability to sell, support, or even switch off access to cutting-edge AI now turns on compliance, not just tech upgrades. For enterprise customers and developers, it means new friction and risk, but also a clearer regulatory playbook taking shape.

But here’s the catch: while Anthropic was scrambling to comply with Washington, it faced a massive threat from the other side—an alleged industrial-scale effort by groups linked to Alibaba to copy Claude’s capabilities via 28.8 million automated queries. This “distillation attack” highlights a stark new reality: the technical moat around these frontier models can be eroded at marginal cost, shifting the battle from model building to relentless defense—identity checks, usage controls, and constant security updates. And with California embracing Anthropic’s Claude for everything from DMV support to Medicaid workflows—even as the Pentagon flagged supply chain risks—the stakes for reliability, pricing, and regulatory credibility are higher than ever.

Featuring reporting from The Washington Post, Bloomberg Law, Politico, Reuters, Tom’s Hardware, and Nikkei Asia.

Powered by Apisod.com