Anthropic close to a deal with the U.S. to lift export controls on its Fable 5 and Mythos 5 models
Anthropic and the Trump administration are about to close a deal that would lift the export restrictions imposed on the company's two most advanced AI models, Fable 5 and Mythos 5, after weeks of security negotiations.
By Bloomberg · June 26, 2026.
Anthropic PBC and the Trump administration are in the final stretch of an agreement that would lift the export controls weighing on the company's two most advanced artificial intelligence models: Fable 5 and Mythos 5. This is according to sources familiar with the matter cited by Bloomberg, who indicate that talks between the two parties have been underway for weeks and are focused on resolving the security concerns that prompted the restrictions.
Commerce Secretary Howard Lutnick is the main driver of the negotiation. According to those same sources, Lutnick is making progress in resolving the security concerns that gave rise to the export controls he himself imposed on Fable 5 and Mythos 5. The lifting of the restrictions would be conditional on the various administration agencies giving their approval, which implies an interagency validation process still underway.
The news confirms that Anthropic's models have been subject to government scrutiny from the standpoint of national security and export control, a trend that has intensified with the frontier models of the leading U.S. AI companies. The imposition of export controls on specific models of a private company is a measure of significant regulatory weight that reflects the growing tension between accelerated technological progress and the federal government's security priorities.
As sector context, the regulation of frontier AI model exports has gained prominence in Washington in recent months. The Trump administration has combined a stance supportive of domestic AI development with a hard line on security to prevent sensitive technology from reaching geopolitical adversaries, especially China. The fact that Anthropic is actively negotiating with the administration to resolve these restrictions suggests that the company believes the controls limit its commercial and competitive capacity internationally.
The article does not detail the technical or security conditions Anthropic would have agreed to meet in order to obtain the lifting of the restrictions, nor does it specify in which markets or applications Fable 5 and Mythos 5 were barred. Since the accessible content of the article is partially limited by Bloomberg's paywall, no further details are available on the scope of the agreement or the expected timeline for its formalization.
From the standpoint of agentic AI, unlocking these models would have direct implications for the developers and companies building agentic systems on Anthropic's infrastructure. The company's frontier models, such as the Claude family, are widely used as reasoning engines in complex agentic pipelines. If Fable 5 and Mythos 5 are Anthropic's most capable models at this time, their availability without export restrictions would broaden international customers' access to advanced agentic capabilities.
Overall, export controls on AI models raise a fundamental debate about the nature of these systems: unlike hardware such as chips, a model's weights can be distributed digitally, which makes export restrictions technically more complex to implement and verify. The agreement in the making between Anthropic and the administration could set a precedent for how these controls are managed for software models in the future.