Anthropic’s team met Trump officials to unwind the Fable export ban after Friday pull

Anthropic met with Trump officials to resolve the export ban on its model Fable, according to POLITICO, and the first-person details point to one thing: the government action was sudden, but the fix is not. Anthropic staff met Monday with senior Trump administration officials for the first in-person sit-downs after the federal government barred non-U.S. users from accessing its newest model, and required the company to pull it entirely to comply. The immediate fight is straightforward to summarize. The government says the model can be used in harmful ways if its safeguards are bypassed. Anthropic says the breach is relatively simple and that other publicly available models can discover the same issues without a “bypass.”

Here is the practical stake: the White House signaled that a quick easing is possible, but only if Anthropic can make its case. A senior White House official said it would likely take longer than a few days to reach a resolution that eases the federal government’s Friday action, which had barred Anthropic from allowing non-U.S. users to access its newest model because of potential security vulnerabilities. The official also left the door open, saying “That’s up to Anthropic.” In other words, this is not a rubber stamp. Anthropic needs to convince the administration that the security problem is minor, not a larger “jailbreak,” and that its safeguards are robust enough for the government’s risk tolerance.

So what happened between the Friday night ban and the Monday meetings? Anthropic reportedly raced to send senior leaders with research and safeguards expertise to Washington, D.C., after multiple hours-long calls over the weekend. Those calls included Anthropic co-founder Tom Brown, Commerce Secretary Howard Lutnick, and National Cyber Director Sean Cairncross, and Anthropic public policy head Sarah Heck was also present. Monday’s in-person meetings, held by Commerce and Cairncross’ office, were described as more technical and led by staff, including Chris Fall, who heads Commerce’s Center for AI Standards and Innovation.

During the Monday session, Anthropic gave a presentation explaining its cybersecurity safeguards in hopes of moving past the restrictions, according to an administration official. Reported participants from Anthropic included Logan Graham, who evaluates and stress-tests models as part of the company’s Frontier Red Team; Dave Orr, its head of safeguards; and Nicholas Carlini, its lead security researcher. The goal for Anthropic is to reduce the government’s perceived risk by grounding the dispute in evidence, not framing. The government’s leverage is that, under the export control, the company cannot simply “patch later.” The model’s release is entangled with access rules, and access rules are entangled with national security judgments.

This matters for more than Anthropic’s product roadmap. The export restriction was described as the most significant escalation yet in the administration’s attempts to regulate the AI industry, which is moving faster than the government’s infrastructure for governing it. It was also characterized as the first time the White House forced a company to remove a model from public access, and it came after a series of “spats” between the administration and Anthropic over ethical restrictions related to potentially catastrophic AI risks. Anthropic has defended its safeguards since the Friday night restrictions were handed down, arguing in a blog post that “These vulnerabilities all appear relatively simple, and we have found that other publicly-available models are able to discover them as well without requiring a bypass.”

What triggered the confrontation? The restriction effectively banned Anthropic from allowing foreign nationals to use Fable 5, and the company said that forced it to pull the model down entirely. The dispute traces back to an earlier disagreement over security concerns with Fable that Amazon raised to the White House late last week. POLITICO had previously reported that Amazon found a way to bypass Anthropic’s guardrails on Fable, and that the discovery was elevated to the White House and the National Security Agency. Anthropic had positioned Fable as a long-awaited model designed to be less powerful than its highly advanced model Mythos, while building in safeguards to prevent prompting for cyberattacks or instructions to create a bioweapon. Mythos, by contrast, had been available only to select companies because of its ability to exploit cyber vulnerabilities faster than any human.

Now the second-order question for leadership teams is not only “Did Amazon’s researchers find an issue?” It is “What kind of issue counts as a control-trigger, and who decides?” Over the weekend, nearly 80 technical experts and CEOs wrote an open letter urging officials to revoke the export control. The letter argued that the actions Amazon researchers got Fable to carry out were a “necessary capability in any model that is intended to write secure code,” and should not be considered an offensive capability. The letter also warned that the administration’s abrupt controls had taken strong models away from defenders, created market uncertainty, and risked America’s AI leadership without any real risk to justify it.

Even within government, there were reportedly differing views on how the rules will play out. A former Trump administration official, Dean Ball, wrote on X that “AI is licensed now, but the requirements change constantly and are always a secret, even to the administration itself, which will discover the rules spontaneously in real time as it reacts to events.” A current administration official warned that the longer the export dispute goes on, the more likely it becomes a de facto licensing regime that AI companies would have to contend with when they release new models, slowing innovation and competition with China. The official said if it does not end quickly with something like a brief “slap on the wrist,” it becomes a huge problem for the entire industry, because it means each model has to ask the government for permission to be released.

That is the competitive stake for founders, investors, and boards across the AI space: this case will likely act as a template for how future model releases are treated. The Monday technical meeting is ostensibly about Fable, but the strategic implication is broader. When a model can be pulled from public access through a security dispute and an export control, the industry stops optimizing only for model performance. It has to optimize for compliance timelines, evidence requirements, and uncertainty. Anthropic’s outcome will not just decide what happens to Fable. It will signal whether the next release is a product moment, or a regulatory negotiation.