Fable 5, 24 hours before the ban: Multiple calls from the White House applying pressure, Anthropic refused to take it down, ultimately faced forced export controls.

Within 24 hours, the US Treasury Secretary, White House Cybersecurity Director, and Commerce Secretary consecutively called Anthropic CEO Dario Amodei, demanding that he proactively take down the newly released AI model Fable 5, but were refused — ultimately, the White House resorted to forceful measures, banning foreign access to the model on national security grounds.

On the 15th, according to media reports citing two government officials and one senior White House official, Amazon CEO Andy Jassy raised concerns about the circumvention of model guardrails with the White House two days after Fable 5 was publicly released, directly triggering the crisis. Afterwards, Treasury Secretary Besent, White House Cyber Director Sean Cairncross, Chief of Staff Susie Wiles, and other senior officials held urgent consultations, then began intensive contact with Amodei. The White House eventually announced export controls on the Fable 5 and Mythos 5 models, prohibiting foreign nationals from using them, and the "net effect" was that all customers could no longer access either model.

This event deeply reveals the fundamental contradictions the White House faces in real-time regulation of rapidly iterating and potentially dangerous AI models, and pushes the already tense relationship between Anthropic and the Trump administration to a new low. Anthropic publicly hit back, calling the export controls "disproportionate" and questioning the transparency and technical basis of the White House's decision process.

Amazon's Report Ignites Crisis

Fable 5 went online publicly this week, with Anthropic defining it as a "Mythos-level model" equipped with safety guardrails for public use. The model had already undergone review by both the US government and the UK AI Safety Institute before launch, and neither side raised clear objections prior to release.

However, just two days after launch, the situation reversed. On Thursday, Amazon CEO Andy Jassy reported the possibility of bypassing the model's guardrails to the White House. According to one source familiar with discussions, Amazon acted in response to a government-initiated request for feedback. Amazon is also an investor in Anthropic.

By Friday morning, the issue had reached the White House's top decision-makers. Besent, Susie Wiles, and other senior officials held an emergency meeting to assess the situation, with Besent remotely joining while headed to a planned event in Houston. The National Security Agency then evaluated Amazon's findings, and officials believed they had "evidence."

Three Calls, Both Sides Deadlocked

The White House attempted to contact Amodei but was told he was unavailable. The White House claimed Amodei was attending a "wellness retreat," which Anthropic spokespersons clearly denied, stating "that is completely false." A person close to Anthropic said the White House contacted them around noon, and Amodei returned the call within an hour and fifteen minutes, with Anthropic also offering to have other executives handle discussions.

Amodei ultimately participated in three rounds of calls, with about six senior White House officials, including Cairncross, Besent, and Commerce Secretary Rutnick. Deputy Secretary for Industry and Security at the Department of Commerce Jeffrey Kessler, White House Office Director Will Scharf, Deputy Chief of Staff Richard Walters, and Presidential Policy Assistant Walker Barrett also joined some of the calls.

In the calls, Amodei defended Fable 5's guardrail system, arguing that the circumvention case was specific and differed from fully disabling all safety restrictions, or "jailbreaking," so the risk was not equivalent. Besent and Cairncross did not accept this reasoning. According to media reports citing senior White House officials, Besent told Amodei directly during the call that he was making a "wrong decision." The White House demanded Anthropic proactively take down the model and work with the government to fix vulnerabilities, but Amodei only requested more time and information, never agreeing to remove the model.

Anthropic: Guardrail System Misinterpreted, Flaws in Decision Process

After Fable 5 was subjected to export controls, Anthropic rebutted in a blog post, saying, "No testers have found a universal jailbreak method—that is, a means to broadly bypass model security measures or unlock all capabilities," and pointed out that completely eliminating jailbreaks is currently infeasible for any company. Anthropic also stated their security system is "strong, with many users even complaining that the restrictions are too broad."

On questioning the decision-making procedure, Anthropic took a tougher stance. According to a source close to the company, the White House gave only 90 minutes for the model to be taken down and "did not provide any specifics about the actual threat." There was never any "request for negotiation," only a 90-minute deadline announcement.

In the statement, Anthropic said the company is following government orders but believes the measures are "disproportionate," asserting that "the government should have the power to prevent unsafe deployments, but this power should be exercised through transparent, fair, clear, and technically-based legal processes, which this action does not meet." Another source close to Anthropic pointed out that before Fable 5's release, the government had communicated many times with Anthropic, and had never opposed the model's launch in these meetings, which contradicts the situation described by the White House afterwards.

Long-standing Tensions, Former White House Official Rarely Supports Regulation Decision

This confrontation was not accidental. On March 3rd this year, Defense Secretary Pete Hegseth and the Pentagon classified Anthropic as a "supply chain risk," citing the company's refusal to allow its AI tools to be used for large-scale domestic surveillance and autonomous weapons, pushing tensions between the two sides to unprecedented heights. Some government officials had previously accused Anthropic of left-wing political bias, and believed the company's advocacy of strict regulation and warnings about mass employment impact were causing panic.

However, former White House AI policy chief David Sacks, who had always opposed regulation, publicly supported imposing export controls on Anthropic this time. He posted on X that he believed this "jailbreak" was neither simple nor minor, and the export controls were not intended to impose broader control on the entire industry.

"The government's hope now is that Anthropic will fix the security issues, the export control will be lifted, and Fable will be reopened to the public," Sacks wrote. "The government wants this to happen as soon as possible. Frankly, the government is puzzled—Anthropic previously claimed security was its highest priority, but is now unwilling to cooperate with safety requests." Sacks also stated that previous disputes between the government and Anthropic are independent of this export control decision.

A White House official who requested anonymity said that the White House was impressed by Anthropic's use of nuclear analogies to describe its technology's dangers, but was "puzzled" by Amodei's refusal to take down the model after known security vulnerabilities. "Export control was the last resort after pleading for their cooperation for hours," the official said. "It was not the course we wanted, but we had no choice."

Risk Disclosure and DisclaimerThe market has risks; invest with caution. This article does not constitute individual investment advice nor does it take into account specific users' investment goals, financial situations, or needs. Users should consider whether any opinions, views, or conclusions in this article suit their own particular circumstances. Investing accordingly is at your own risk.