Anthropic Staff Heads to D.C. Amid Export Control Tension
In a bid to address the White House’s recent actions regarding its AI models, some senior staff members from Anthropic traveled to Washington, D.C. over the weekend. A source close to the company informed The Post that staff had been in virtual discussions with the Trump administration since the White House imposed export controls on the models known as Mythos and Fables last week.
It appears that Anthropic is eager to collaborate with the administration to resolve the ongoing conflict. The company’s team planned to meet with officials from the Commerce Department on Monday as part of these discussions.
The export controls were introduced on Friday after receiving insights from Amazon CEO Andy Jassy, who raised concerns about potential national security threats linked to the guardrails of these AI models. In response, Anthropic took the drastic step of removing its models from online access, arguing that it was necessary to ensure compliance, despite disagreeing with the White House’s decision.
The White House had reached out to the Commerce Department for a comment, though no immediate response was provided.
According to reports, White House officials expressed frustration that Anthropic minimized the security issues associated with Fables, referring to them as a mere “narrow jailbreak possibility”—an issue not exclusive to their models. This comes in contrast to CEO Dario Amodei’s frequent assurances that AI safety is a top priority for the company.
A senior official noted that before the administration took action, the White House was alerted by not just Amazon, but also by “nearly half a dozen” companies raising concerns about Fables.
Commerce Secretary Howard Lutnick, along with White House Cyber Director Sean Cairncross, has been engaged with Anthropic’s team regarding these security matters.
Meanwhile, several high-profile cybersecurity officials have come to Anthropic’s defense, urging the Trump administration to reconsider its export control policies. In a letter to Lutnick and Cairncross, they argued that this action lacked justification, noting it deprived defenders of access to superior models, created market uncertainties, and endangered U.S. leadership in AI.
The letter’s signatories included over 40 individuals, among them cybersecurity leaders from major firms like Adobe and Zoom, along with Alex Stamos, former chief security officer of Facebook.
Experts have weighed in, suggesting that the “jailbreak” technology identified by Amazon and others is not a unique vulnerability of Fables, but rather one that could easily be replicated in various models, including Chinese ones like GPT-5.5, Opus, Sonnet, and Kimi 2.7.
Adam Thierer, a senior policy analyst at the R Street Institute, posted on X to describe the Trump administration’s actions as a significant escalation in the politicization of AI and the centralization of control over advanced computing in the U.S.
Recent conflicts with the White House have posed challenges for Anthropic, which is currently in a legal battle over the Pentagon’s classification of its AI models as a supply chain risk.
As Anthropic races to go public ahead of competitors like OpenAI, this latest development complicates their situation, especially with their primary product now inaccessible to consumers. Amodei and other leading tech CEOs are expected to attend the G7 meeting in Paris this week, which adds another layer of complexity to the unfolding scenario.
