Anthropic Models Remain Offline After Unclear Export Control Order

Regulatory limbo for advanced AI models

Anthropic's most advanced AI models have been offline for nearly a week following an export control directive from the Trump administration, and the company still cannot distribute Claude Mythos or Fable 5. Despite ongoing negotiations between Anthropic and the White House, the two parties remain at an impasse—and the specific violation that triggered the action remains unclear.

The situation highlights a fundamental challenge in AI governance: the government is effectively creating enforcement precedents without established rules. When asked what Anthropic did wrong, the answer depends on whom you ask, according to reporting first published by WIRED.

Why it matters

This case reveals how AI regulation is being improvised in real time at the highest levels of government. When leading AI companies cannot determine what compliance looks like—even after direct negotiations with White House officials—it creates uncertainty that affects product development, international competitiveness, and the broader AI industry's ability to plan. The lack of clear standards also raises questions about due process and whether enforcement actions are based on technical assessments or shifting political considerations.

No clear path forward

The Trump administration issued the export control directive that forced Anthropic to take its models offline, but days of high-level discussions have not resolved the standoff. The absence of a clear explanation for what triggered the action leaves Anthropic without a roadmap for bringing its models back into distribution.

This regulatory ambiguity comes despite the fact that Anthropic is one of the world's leading AI laboratories, with significant resources to navigate compliance requirements. If a well-funded company with direct access to White House officials cannot determine the rules, smaller organizations face an even more challenging environment.

Broader implications for AI governance

The Anthropic case is not an isolated incident but part of a broader pattern of ad hoc AI policymaking. The Trump administration previously eliminated an executive order designed to regulate AI, leaving companies and officials scrambling to understand what framework, if any, governs advanced AI development and deployment.

Export controls have traditionally been used to restrict sensitive technology from reaching adversaries, but applying them to AI models raises novel questions about what constitutes a violation, how capabilities are assessed, and what remediation looks like. The current situation suggests these questions are being answered through enforcement actions rather than through clearly articulated policy.

The details of this ongoing dispute were first reported by WIRED.