Anthropic and the Trump Administration Meet Over AI Cybersecurity Dispute
Anthropic has sent senior technical staff to Washington, D.C., to meet directly with Trump administration officials. The meeting, scheduled at the US Commerce Department, follows an unprecedented export-control directive issued on Friday that orders Anthropic to suspend access to its Fable 5 and Mythos 5 AI models.
The Anthropic–Trump administration meeting marks one of the most significant clashes between a frontier AI company and the US government. A person familiar with the situation said Anthropic’s senior technical staff has held daily virtual calls with White House officials since the directive was issued. They also travelled to Washington on Monday for an in-person meeting. The names of the participants have not been disclosed. The discussions remain confidential.
The government cited national security concerns and ordered Anthropic to suspend access to Fable 5 and Mythos 5 for all foreign nationals. As a result, Anthropic disabled the models for all customers to ensure compliance. The dispute centres on an alleged jailbreak of Fable 5. On June 13, senior Trump adviser David Sacks said the administration had received warnings about the issue. He added that CEO Dario Amodei considered the risk minor and chose not to delay the release.
Meanwhile, a source close to the White House said Amazon CEO Andy Jassy was among several tech leaders who warned officials about cybersecurity risks in Anthropic’s latest models. However, Anthropic said its review found that the jailbreak relied on minor and previously known vulnerabilities. The company noted that other publicly available models, including OpenAI’s GPT-5.5, could also discover them. It also said the government sent the directive at 5:21 p.m. ET on Friday without any prior warning.
Sources familiar with the matter said the talks focus on jailbreak vulnerabilities, safety standards, and the basis for the export controls. In response, Anthropic called the action a misunderstanding and said it was working to restore access. The company argued that applying the same standard across the industry would effectively halt new frontier AI releases. Meanwhile, less powerful Claude models, including Claude Opus 4.8, remain available. More broadly, the outcome could shape how the US government oversees future AI systems. Therefore, the industry is watching the discussions closely.
