Editor's Note: Anthropic had just released its Fable 5 model a few days ago when it was suddenly forced to take it offline. The Trump administration cited national security concerns to restrict foreign nationals' access to Anthropic's Fable 5 and Mythos 5 advanced AI models. Due to the difficulty of accurately distinguishing user identities and access rights, Anthropic ultimately chose to disable the relevant models for all customers.
Behind this sudden takedown was a high-pressure standoff between the White House and Anthropic within a 24-hour period. Regarding whether the security barrier around Fable 5 could be circumvented, Anthropic CEO Dario Amodei engaged in multiple tense calls with senior government officials from the Treasury Department, Commerce Department, White House IT leadership, and others. The government believed the models could be used to identify software vulnerabilities, posing a national security risk; Anthropic argued that the issues in question were not a "general jailbreak" and criticized the government's actions for lacking a transparent, clear, and technically grounded legal process.
The key to this event is not just a model being taken offline or a breakdown in communication between the White House and an AI company, but the redefinition of advanced AI models as a potential national security asset. As model capabilities enter sensitive scenarios such as cybersecurity, vulnerability discovery, and intelligence analysis, the government is no longer satisfied with corporate self-assessment and voluntary reviews but is beginning to use more stringent policy tools like export controls to directly influence model release cadence.
Moreover, it is ironic that Anthropic has been one of the most vocal advocates for regulation within the AI industry but has now become the target of regulatory actions. The government believes the company did not respond seriously enough to security vulnerabilities, while Anthropic considers the White House's response to be excessive. Behind the controversy lies a new issue that the AI industry must face in the future: Who has the authority to determine if a model is sufficiently secure? Is self-certification by companies enough? How will rapid intervention in the name of national security change the compliance costs, release cadence, and global competitive landscape of AI companies?
The Anthropic incident could mark a turning point in the regulation of advanced AI. 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 industry order.
The following is the original text:
Before the Trump administration imposed a comprehensive export control on Anthropic, White House senior officials were involved in a 24-hour emergency mediation, trying to persuade the AI company to voluntarily take down a newly released model. Several government officials believed that the model might pose a national security risk.
According to two government officials and a White House senior official, during this turmoil, Anthropic CEO Dario Amodei had tense phone conversations with several government officials, including U.S. Treasury Secretary Scott Bessent and White House Chief of Networks Sean Cairncross. As discussions of this matter were not authorized to be public, the individuals involved requested anonymity.
This incident also brought to the forefront an increasingly urgent issue: faced with rapidly iterating advanced AI models pushing the boundaries of capability, how should the White House draw the line between innovation and security?
These details of the conversations have never been reported before.
Eve of Model Takedown: Security Controversy
Following the government's export control implementation, Anthropic was forced to take down its new model, Fable. The model had been publicly available for only a few days. Anthropic had previously assured the public of the model's safety and controllability, but shortly after the release, government officials began reevaluating whether its security barriers were as robust as the company claimed.
Two government officials and a White House senior official said that on Thursday, just two days after Fable's public release, Amazon CEO Andy Jassy expressed concerns to the White House, suggesting that there might be risks of bypassing the model's security barriers.
A person familiar with Amazon's related communications stated that Amazon was responding to government requests to provide feedback on potential risks.
By early Friday morning, the issue had escalated to the highest levels of the White House.
According to government officials and a White House senior 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 plan. One insider mentioned that Bessent was en route to Houston for a scheduled public event and thus participated remotely.
After the meeting, the government attempted to reach Amodei but was informed that he was temporarily unavailable to take calls due to participating in a health retreat, according to a government official and a White House senior official.
Anthropic spokesperson denied the claim, stating, "This is completely untrue."
A source close to Anthropic said the government first attempted to reach Amodei around midday, and within 1 hour and 15 minutes, Amodei was on the phone with senior officials. The source mentioned that during a period when Amodei was temporarily unavailable, Anthropic offered to have other executives communicate on their behalf.
After finally reaching Amodei, he participated in three separate calls. White House senior officials and a government official noted that around half a dozen senior government officials were involved in the calls, including Cairncross, Bessent, and Secretary of Commerce Howard Lutnick.
White House senior officials stated that other White House aides and government officials also participated in parts of the calls, including Deputy Secretary of Commerce for Industry and Security Jeffrey Kessler, White House Staff Secretary Will Scharf, White House Deputy Chief of Staff Richard Walters, and Presidential Policy Assistant Walker Barrett.
During the calls, Amodei attempted to clarify what he believed were misunderstandings. He addressed the government's security concerns, defended Anthropic's safeguards, and emphasized that the bypasses that had occurred were specific scenario issues and not equivalent to a broader "jailbreak"—where the model would be entirely free of Anthropic's security constraints.
In a blog post released after the implementation of export controls, Anthropic stated, "There is currently no tester who has found a universal jailbreak—a way to broadly bypass model security measures and unlock significant networking capabilities. The company also expressed that completely eliminating all jailbreaking attempts is currently not feasible for Anthropic or any other AI company.
Anthropic also defended its security systems, stating that its safeguards are "so strong that many users complain they are overly restrictive."
However, Cairncross and Bessent were not convinced by Amodei's explanation. A White House official mentioned that Amazon's findings had been submitted for review by the National Security Agency, which deemed the materials sufficient to constitute "evidence."
According to White House senior officials and two government officials, the government urged Anthropic to voluntarily take down the model and work with the government to address the vulnerabilities. Amodei requested more time and information but did not commit to taking down the model. A White House official stated that during one call, Bessent directly told Amodei that he was making a "bad decision."
Regulatory Crackdown: Fable 5 and Mythos 5 Face Restrictions
Soon after the call ended, the Trump administration imposed export controls on the Fable 5 and Mythos 5 models. Anthropic stated that the government invoked national security powers, prohibiting foreign nationals from using these models. The company said the "practical effect" of this order was that, to ensure compliance, Anthropic had to "suddenly disable" all client access to the relevant models.
A senior White House official said, "We pleaded with them for hours; export controls were a last resort. This is not something we wanted to do, but we had no choice."
In response to the report, a source close to Anthropic countered that the company was not given the option of so-called "voluntary cooperation."
The source said, "The White House only gave a 90-minute takedown notice for the models, and provided no details about the actual threat. There was no plea, no request to work with them; just a declared 90-minute deadline."
A senior White House official stated that White House officials had heard Amodei liken the danger of Anthropic's technology to a nuclear weapon. Therefore, when the CEO, who has always advocated for AI safety regulation, refused to take down the system to address known security vulnerabilities, government officials found it hard to understand.
Anthropic has long been seen as one of the most vocal advocates for regulation in the AI industry. The company has repeatedly argued that a regulatory framework should address the global security risks and job displacement that could arise from the rapid advancement of AI.
Three individuals familiar with the government's thinking said that Amazon was not the only company that had expressed concerns to the government.
One of them said, "The crux of the issue is that Anthropic did not take this matter seriously enough. If Anthropic had taken it seriously, instead of dismissing it as an isolated incident and proactively fixing or pausing access, this would never have escalated to this point."
Another source close to Anthropic countered that the so-called "jailbreak" did not mean a failure of Fable 5's security system, pointing out that the company had communicated with the government before releasing Fable. The source said that in multiple rounds of communication, the government did not oppose the release of Fable.
Following the government's implementation of export controls, Anthropic stated in a blog post that it would comply with the government's directive but believed that this move was an overreaction.
Anthropic stated: "As we have previously stated publicly, we believe that governments should have the authority to prevent unsafe deployments, but this should be done through a transparent, fair, clear, and technically grounded legal process. This action does not align with these principles."
A White House official, speaking on the condition of anonymity, said that while innovation remains a "top priority for the White House, we must also prioritize security."
Amazon, in a statement, refused to disclose specific details of its communications with the government. A spokesperson for Amazon said, "It is not uncommon for the government to seek our views on potential security risks. When such situations arise, we do not disclose details of the discussions."
In early April of this year, Anthropic announced that its latest powerful model, Mythos, would only be made available to a select group of tech and cybersecurity companies to help them test software vulnerabilities. Anthropic stated at the time that due to the model's excessive capabilities, if it fell into the wrong hands, it could have serious consequences, thus the release needed to be restricted.
The debut of Mythos triggered a series of meetings between Amodei and senior White House officials. Both sides described these meetings as productive. Subsequently, there were multiple rounds of discussions on how to regulate advanced AI models, leading to a recent executive order requiring companies to voluntarily submit models for government review before widespread deployment.
The publicly released Fable 5 this week was described by Anthropic as a "Mythos-level model," but with additional security measures to make it available to the general public. The model had undergone reviews by the U.S. government and the UK AI Security Institute.
However, after so-called security flaws were disclosed, multiple government officials believe that the model must be taken down immediately.
On Saturday morning, former White House AI Lead, and a staunch opponent of regulation, David Sacks, posted on X in support of the government's export controls on Anthropic.
Sacks stated that he did not believe this "jailbreak" was simple or insignificant, nor did he think export controls were the government's attempt to broadly control the AI industry. Sacks wrote, "The government now wants Anthropic to fix the security issues, then lift the export controls, and have Fable return to public release status. The government wants this to happen quickly. Frankly, what's confusing to the government is that Anthropic has always said security requests were its top priority, but now is unwilling to cooperate."
Sacks and other government officials have previously criticized Anthropic, accusing it of having a left-wing political bias and viewing the company as fearmongering due to its advocacy for stronger industry regulation and warnings of large-scale job displacement.
Earlier this year, Defense Secretary Pete Hegseth escalated the government's disagreement with Anthropic to unprecedented levels. On March 3, the Pentagon identified the company as a supply chain risk after Anthropic refused to allow its AI tools to be used for large-scale domestic surveillance and autonomous weapon systems.
On Saturday, Sacks stated that the existing dispute between the government and Anthropic is separate from this export control decision. Sacks wrote, "The government values Anthropic's technical capabilities and sees this issue as serious but easily addressable. The ball is now in Anthropic's court."
Welcome to join the official BlockBeats community:
Telegram Subscription Group: https://t.me/theblockbeats
Telegram Discussion Group: https://t.me/BlockBeats_App
Official Twitter Account: https://twitter.com/BlockBeatsAsia
