Inside Anthropic's AI Export Ban: 14 Days That Could Redefine Tech Law

The U.S. government’s unprecedented intervention into Anthropic’s operations—imposing export controls on two AI models via a Commerce Department letter—raises profound legal questions about the scope of executive authority over cutting-edge technology. With a resolution now in sight, the negotiations are crafting a de facto regulatory framework that could define how future administrations police AI, from the limits of trade law to the standards of evidence required for such drastic action.

Anthropic, the AI research company, is edging closer to a resolution with the Trump administration that would lift export controls on its most powerful artificial intelligence models, Fable 5 and Mythos 5, after a surprise government intervention two weeks ago that barred foreign nationals from accessing the systems. The negotiations, led by Commerce Secretary Howard Lutnick and involving top Anthropic executives like co-founder Tom Brown, mark an unprecedented direct clash between the U.S. government and a private AI lab over the security and deployment of cutting-edge models. The restrictions, delivered abruptly in a letter from Lutnick, forced Anthropic to disable global access to both systems—a drastic step that highlighted the escalatory potential of AI-related national security policy.

“

The negotiations, led by Commerce Secretary Howard Lutnick and involving top Anthropic executives like co-founder Tom Brown, mark an unprecedented direct clash between the U.S.

The standoff originates from concerns that the safety guardrails built into Fable 5 and Mythos 5 could be circumvented by foreign actors, potentially exposing sensitive capabilities or enabling malicious use. The government's action, the most significant intervention of its kind into an AI venture's operations, leveraged trade regulations typically reserved for dual-use technology like advanced semiconductors. It reflects a growing belief within the administration that frontier AI models require proactive oversight, especially as they approach capabilities that could disrupt economic or military balances. The letter gave no explicit timeline, but the immediate disabling of access signaled the severity of the matter and kicked off intensive behind-the-scenes talks.

For Anthropic, the temporary loss of access to its top-tier systems represented not only a commercial setback but also a reputational challenge. The company, known for its safety-first approach to AI, found itself at the center of a national security debate it had sought to preempt through voluntary safety measures. The negotiations, therefore, are as much about proving the robustness of its internal safeguards as they are about defining the rules of engagement between government and industry. The involvement of co-founder Tom Brown in direct high-level meetings underscores the seriousness with which Anthropic is approaching the situation—and the political capital it is willing to spend to regain unrestricted access.

A U.S. official confirmed that the Commerce Department is working toward a resolution and striving to align the entire administration on any decision, indicating that a multi-agency review is underway. That process could involve the White House, the Defense Department, and intelligence agencies, each weighing in on the potential risks. Until all have signed off, the restrictions remain in force. The broad administrative consensus requirement could slow a final deal, but sources suggest that progress is being made and that the curbs could be dropped relatively quickly once all security concerns are addressed.