Fable5被迫下线前24小时，白宫发生了什么？

Original title: Inside the whirlwind 24 hours that led the White House to slap export controls on Anthropic

Original authors: Sophia Cai and Cheyenne Haslett, Politico

Original translation: Peggy

Editor's note: The Fable 5 model, which Anthropic had just released a few days earlier, was suddenly taken offline. The cause was the Trump administration citing national security concerns to restrict foreign citizens' access to Anthropic's two advanced AI models, Fable 5 and Mythos 5. Unable to precisely distinguish user identity and access permissions, Anthropic ultimately chose to disable the relevant models for all customers.

Behind this sudden takedown was a high-pressure 24-hour standoff between the White House and Anthropic. Centered on whether the safety guardrails of Fable 5 could be bypassed, Anthropic CEO Dario Amodei engaged in several tense phone calls with senior government officials, including the Treasury Secretary, Commerce Secretary, and the White House cyber director. The government believed the model could be used to identify software vulnerabilities, posing a national security risk. Anthropic, however, emphasized that the issue was not a "general jailbreak" and criticized the government's lack of transparent, clear, and technically grounded legal procedures.

The key to this incident is not just a model being taken offline, nor merely a breakdown in communication between the White House and an AI company. It marks a redefinition of advanced AI models as potential national security assets. When model capabilities enter sensitive domains like cybersecurity, vulnerability exploitation, and intelligence analysis, the government is no longer satisfied with corporate self-assessment and voluntary reviews. It has begun to employ tougher policy tools like export controls, directly intervening in the pace of model releases.

Ironically, Anthropic has been one of the most vocal advocates for regulation within the AI industry, yet it has now become the target of the regulatory hammer. The government believes it didn't take security vulnerabilities seriously enough, while Anthropic considers the White House's response an overreaction. At the heart of this dispute is a new question the AI industry must face: Who has the right to judge whether a model is safe enough? Is corporate self-certification of safety still sufficient? And how will rapid interventions under the guise of national security change compliance costs, release timelines, and the global competitive landscape for AI companies?

The Anthropic incident could become a watershed moment for advanced AI regulation. In the future, the release of cutting-edge models will no longer be just a product launch, but more akin to a high-risk deployment involving national security, geopolitics, capital markets, and industrial order.

The following is the original text:

Before the Trump administration imposed sweeping export controls on Anthropic, senior White House officials engaged in a 24-hour emergency mediation effort, attempting to persuade the AI company to voluntarily take down a newly released model. Multiple government officials believed the model posed a potential national security risk.

According to two government officials and one senior White House official, during this upheaval, Anthropic CEO Dario Amodei had several tense phone calls with high-ranking government officials, including U.S. Treasury Secretary Scott Bessent and White House Cyber Director Sean Cairncross. These individuals spoke on condition of anonymity as they were not authorized to discuss the matter publicly.

This incident also brings an increasingly pressing issue to the forefront: How should the White House draw the line between innovation and security when faced with rapidly iterating advanced AI models whose capabilities are constantly expanding?

These details of the calls have never been reported before.

The Night Before the Model Takedown: Security Dispute

Following the government's imposition of export controls, Anthropic was forced to take down its new model, Fable, which had been publicly available for only a few days. Anthropic had previously assured the public the model was safe and manageable, but shortly after its release, senior government officials began reassessing whether its safety guardrails were as robust as the company claimed.

Two government officials and one senior White House official said that on Thursday, two days after Fable's public release, Amazon CEO Andy Jassy expressed concerns to the White House that the model's safety guardrails might be bypassed.

A person familiar with Amazon's communications said the company was responding to a government request for feedback on potential risks.

By Friday morning, the issue had escalated to the highest levels of the White House.

According to the government officials and the senior White House official, Bessent, Cairncross, White House Chief of Staff Susie Wiles, and other senior officials held a meeting to discuss the model and the government's response. One source said Bessent participated remotely as he was en route to Houston for a scheduled public event.

After the meeting, the government tried to contact Amodei but was told he was temporarily unavailable due to attending a health retreat, according to one government official and the senior White House official.

An Anthropic spokesperson denied this, stating: "This is completely untrue."

A person close to Anthropic said the government first requested to contact Amodei around noon, and Amodei was on the phone with senior officials within one hour and fifteen minutes. This person stated that during the time Amodei was temporarily unavailable, Anthropic offered to have other executives communicate on his behalf.

Once the government finally reached Amodei, he participated in three phone calls. The senior White House official and a government official said about half a dozen senior government officials were on the calls, including Cairncross, Bessent, and Commerce Secretary Howard Lutnick.

The senior White House official said other White House staff and government officials joined parts of the calls, including Commerce Department Under Secretary for Industry and Security Jeffrey Kessler, White House Cabinet Secretary Will Scharf, White House Deputy Chief of Staff Richard Walters, and Presidential Policy Advisor Walker Barrett.

During the calls, Amodei tried to clarify what he believed were misunderstandings. He pushed back against the government's safety concerns, defended Anthropic's safeguards, and emphasized that the methods used to bypass them were specific scenarios and did not constitute a broader "jailbreak"—allowing the model to completely escape Anthropic's safety guardrails.

In a blog post published after the export controls were implemented, Anthropic stated: "To date, no testers have found a general jailbreak method—one that could broadly bypass the model's safety measures and unlock a wide range of cyber capabilities." The company also said it's currently unrealistic for Anthropic or any other AI company to completely prevent all jailbreak attempts.

Anthropic also defended its safety systems, stating its guardrails are "so strong that many users complain they are overly broad."

However, Cairncross and Bessent were not convinced by Amodei's explanations. A White House official said Amazon's findings were submitted to the National Security Agency for review, which considered the material sufficient as "evidence."

According to the senior White House official and two government officials, the government urged Anthropic to voluntarily take down the model and coordinate with the government to fix the vulnerabilities. Amodei asked for more time and information but did not commit to taking the model down. The senior White House official stated that during one of the calls, Bessent directly told Amodei he was making a "bad decision."

The Regulatory Hammer Falls: Fable 5 and Mythos 5 Face Controls

Shortly after the calls concluded, the Trump administration imposed export controls on the Fable 5 and Mythos 5 models. Anthropic stated the government invoked national security authority to ban foreign citizens from using these models. The company said the "practical effect" of the order was that, to ensure compliance, Anthropic had to "abruptly disable" all customer access to the relevant models.

A senior White House official said: "We pleaded with them to work with us for hours. The export controls were a last resort. It's not something we wanted to do, but we felt we had no choice."

After the report was published, a person close to Anthropic refuted this, claiming the company was not given an option for "voluntary cooperation."

This person said: "The White House gave us 90 minutes to take the model down and provided no details about the actual threat. There was never any pleading, nor a request to work with us. It was just an announced 90-minute deadline."

A senior White House official said White House officials had heard Amodei compare the dangers of Anthropic's technology to a nuclear bomb. Therefore, when the CEO who had long advocated for AI safety regulation was unwilling to take down a system to fix known security vulnerabilities, government officials found it incomprehensible.

Anthropic has long been seen as one of the AI industry's most vocal proponents of regulation. The company has repeatedly argued that a regulatory framework is necessary to address the potential global security risks and employment impacts of rapid AI development.

Three people familiar with the government's thinking said Amazon was not the only company to raise concerns with the government.

One of them said: "The core of the problem is that Anthropic didn't take this matter seriously enough. If Anthropic had taken it seriously, instead of downplaying it as an isolated incident, and proactively fixed it or paused access, this never would have escalated to this point."

Another person close to Anthropic countered, arguing that the so-called 'jailbreak' didn't mean Fable 5's safety systems were ineffective, and noted the company had communicated with the government before releasing Fable. This person stated that in multiple rounds of communication, the government did not object to Fable's release.

After the government imposed export controls, Anthropic stated in a blog post that it would comply with the government directive, but also believed the measure was an overreaction.

Anthropic said: "As we have stated publicly, we believe the government should have the authority to prevent unsafe deployments. However, this should be based on a transparent, fair, clear, and technically grounded legal process. This action does not align with these principles."

A White House official, speaking on condition of anonymity, said innovation remains the White House's "top priority, but we must also prioritize safety."

Amazon declined to disclose details of its communication with the government in a statement. An Amazon spokesperson said: "It is not unusual for the government to seek our input on potential security risks. When such situations arise, we do not disclose the details of those discussions."

In early April of this year, Anthropic announced that its latest powerful model, Mythos, would only be made available to a small number of tech and cybersecurity companies to help them test software vulnerabilities. At the time, Anthropic stated that due to the model's immense power, it could cause severe consequences if it fell into the wrong hands, hence the need to restrict its release.

The unveiling of Mythos led to a series of meetings between Amodei and senior White House officials. Both sides described these meetings as productive. Subsequently, multiple rounds of discussions on how to regulate advanced AI models took place, culminating in a recent executive order requiring companies to voluntarily submit advanced models for government review before widespread deployment.

Fable 5, released publicly this week, was described by Anthropic as a "Mythos-class model" but with additional safety measures allowing it to be opened to general users. The model had been reviewed by both the U.S. government and the UK AI Safety Institute.

However, after the alleged safety flaws were disclosed, multiple government officials believed the model needed to be taken down immediately.

On Saturday morning, former White House AI director David Sacks, a staunch opponent of regulation, posted on X in support of the government's export controls on Anthropic.

Sacks stated he didn't believe the 'jailbreak' was simple or trivial, nor did he think the export controls were an attempt by the government to broadly control the AI industry. Sacks wrote: "The government now wants Anthropic to fix the security issues, then lift the export controls, and return Fable to public release. The government wants this to happen as quickly as possible. Frankly, the government is baffled that Anthropic, which previously said safety requests were its highest priority, is now unwilling to cooperate."

Sacks and other officials within the government have previously criticized Anthropic, accusing it of left-wing political bias and of fear-mongering by advocating for stronger industry regulation and warning of massive job displacement.

Earlier this year, Defense Secretary Pete Hegseth and the Pentagon elevated the rift between the government and Anthropic to an unprecedented level. On March 3, because Anthropic refused to allow its AI tools to be used for mass domestic surveillance and autonomous weapons systems, the Pentagon listed the company as a supply chain risk.

On Saturday, Sacks said the pre-existing dispute between the government and Anthropic was separate from the export control decision. Sacks wrote: "The government values Anthropic's technical capabilities and believes this issue, while serious, should be easily resolvable. The ball is now in Anthropic's court."

Original Link