Lead: Just days after its release, Anthropic's Fable 5 model was abruptly taken offline. The trigger? The Trump administration invoked national security concerns to restrict foreign access to two advanced AI models from Anthropic—Fable 5 and Mythos 5. Unable to reliably distinguish user identity and access permissions, Anthropic ultimately opted to disable the models for all customers.

This sudden takedown stems from a high-pressure 24-hour standoff between the White House and Anthropic. Amid intense phone calls involving Anthropic CEO Dario Amodei and senior government officials including the Treasury Secretary, Commerce Secretary, and the White House Cybersecurity Coordinator, debate raged over whether Fable 5’s safety safeguards could be bypassed. The government feared the model might be exploited to identify software vulnerabilities, posing a national security threat. In contrast, Anthropic emphasized that the observed exploits were not "general jailbreaks" and criticized the government’s actions as lacking transparency, clarity, and a legally grounded technical assessment.

The core significance of this incident extends beyond a single model removal or breakdown in communication between a tech firm and the government—it marks a pivotal shift in how advanced AI models are being redefined as potential national security assets. As model capabilities increasingly intersect with sensitive domains like cybersecurity, vulnerability discovery, and intelligence analysis, governments are no longer content with self-assessments or voluntary compliance. Instead, they are resorting to hardline tools such as export controls to directly intervene in model deployment timelines.

Ironically, Anthropic has long been one of the most vocal advocates for AI regulation within the industry. Now, it finds itself on the receiving end of regulatory enforcement. Government officials argue that Anthropic responded too lightly to security vulnerabilities; Anthropic counters that the White House overreacted. Underlying this clash lies a fundamental question facing the AI industry moving forward: Who holds authority to determine whether a model is sufficiently safe? Is corporate self-certification enough? And how will rapid, national-security-based interventions reshape AI firms’ compliance costs, release cycles, and global competitive positioning?

The Anthropic incident may mark a watershed moment in the governance of frontier AI. Going forward, deploying cutting-edge models will no longer be merely a product launch event but a high-stakes operation involving national security, geopolitics, capital markets, and industrial order.

Below is the original text:

Before the Trump administration imposed comprehensive export controls on Anthropic, senior White House officials spent 24 hours in emergency negotiations attempting to persuade the AI company to voluntarily pull down a newly launched 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 crisis, Anthropic CEO Dario Amodei engaged in several tense phone calls with top-level government figures, including U.S. Treasury Secretary Scott Bessent and White House Cybersecurity Coordinator Sean Cairncross. Due to restrictions on public discussion, all sources requested anonymity.

This episode brings to the forefront an increasingly urgent question: How should the White House draw the line between innovation and security in the face of rapidly evolving, boundary-expanding advanced AI models?

These call details have never been reported before.

After export controls were enacted, Anthropic was forced to take down its new model, Fable. The model had been publicly available for only a few days. While Anthropic had previously assured the public that the model was secure and controllable, government leaders began reassessing the robustness of its safety barriers shortly after launch.

Two government officials and one senior White House official said that on Thursday—just two days after Fable’s public release—Amazon CEO Andy Jassy expressed concerns to the White House, warning that the model’s safety protections might be vulnerable to circumvention.

A source familiar with Amazon’s internal communications said Amazon 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 government and senior White House officials, Bessent, Cairncross, White House Chief of Staff Susie Wiles, and other senior officials convened a meeting to discuss the model and potential government responses. One informed source said Bessent was en route to Houston for a scheduled public event and joined remotely.

Following the meeting, government officials attempted to contact Amodei, only to be told he was temporarily unreachable due to participation in a health retreat. This account was confirmed by a government official and a senior White House official.

Anthropic spokesperson denied the claim, stating: “This is entirely false.”

A source close to Anthropic countered that government officials first tried to reach Amodei around noon, and he connected with senior officials within 1 hour and 15 minutes. The source added that during the period when Amodei was unavailable, Anthropic offered to facilitate communication through other executives.

After finally reaching Amodei, he participated in three separate calls. According to a senior White House official and a government official, approximately half a dozen senior government officials took part, including Cairncross, Bessent, and Commerce Secretary Howard Lutnick.

The senior White House official noted that additional aides and officials also joined portions of the conversations, including Deputy Under Secretary for Industrial and Security Affairs Jeffrey Kessler at the Department of Commerce, White House Clerk Will Scharf, Deputy Chief of Staff Richard Walters, and Presidential Policy Advisor Walker Barrett.

During the calls, Amodei sought to clarify what he viewed as misunderstandings. He disputed the government’s security concerns, defended Anthropic’s protective mechanisms, and stressed that the methods identified so far were context-specific issues—not equivalent to a broad “jailbreak” defined as completely circumventing Anthropic’s safety guardrails.

In a blog post released after the export controls took effect, Anthropic stated: “No testing team has yet found a universal jailbreak method—one capable of broadly bypassing model safeguards and unlocking extensive network capabilities.” The company added that completely eliminating all jailbreak attempts remains unrealistic for Anthropic or any other AI firm.

Anthropic further defended its security systems, noting that their safeguards were “so strong that many users have complained they are overly broad.”

However, Cairncross and Bessent remained unconvinced by Amodei’s explanations. A White House official said Amazon’s findings had been reviewed by the National Security Agency, which deemed the materials sufficient as “evidence.”

According to senior White House officials and two government sources, the government urged Anthropic to proactively remove the model and coordinate on fixing vulnerabilities. Amodei requested more time and information but did not commit to taking the model offline. A senior White House official said that during one call, Bessent explicitly told Amodei he was making a “bad decision.”

Shortly after the calls concluded, the Trump administration imposed export controls on both Fable 5 and Mythos 5. Anthropic stated that the government invoked national security authorities to prohibit foreign citizens from using these models. The company said the “practical effect” of the order was that, to ensure compliance, it had no choice but to “suddenly disable” all customer access to the affected models.

A senior White House official said: “We asked them to cooperate with us for several hours. Export controls were a last resort. It’s not what we wanted to do, but we had no alternative.”

After the report was published, a source close to Anthropic pushed back, asserting the company was never given a genuine “voluntary cooperation” option.

The source said: “The White House gave only a 90-minute deadline to take down the model—and provided no details about the actual threat. There was no plea for cooperation, no request to work together. Only a declared 90-minute ultimatum.”

A senior White House official said government officials had heard Amodei liken the danger of Anthropic’s technology to that of nuclear weapons. Thus, when the CEO—who has long advocated for AI safety regulation—refused to take down a system to fix known security flaws, government officials found the stance incomprehensible.

Anthropic has long been regarded as one of the most proactive voices calling for AI regulation in the industry. The company has repeatedly argued that a regulatory framework is necessary to address the global security risks and employment disruptions potentially triggered by the rapid advancement of AI.

Three individuals familiar with government thinking said Amazon was not the only company raising concerns with the government.

One said: “The core issue is that Anthropic didn’t treat this seriously enough. If they had taken it seriously instead of downplaying it as an isolated incident and proactively paused access or fixed the issue, this situation would never have reached this point.”

Another source close to Anthropic pushed back, arguing that the term “jailbreak” does not imply a failure of Fable 5’s safety systems and pointed out that Anthropic had maintained communication with the government prior to releasing Fable. The source noted that during multiple rounds of discussions, the government had not objected to the Fable release.

After the export controls were implemented, Anthropic stated in a blog post that it would comply with government directives but considered the move disproportionate.

Anthropic said: “As we’ve publicly stated before, we believe the government should have the authority to prevent unsafe deployments—but this must be done through a transparent, fair, clear, and technically grounded legal process. This action does not meet those principles.”

A government official granted anonymity said: “Innovation remains our top priority, but security must also be a primary concern.”

Amazon declined to disclose specific details of its communications with the government. A spokesperson said: “It’s not uncommon for the government to seek our input on potential security risks. When such situations arise, we do not disclose the details of our discussions.”

At the beginning of April, Anthropic announced that its latest powerful model, Mythos, would be made available only to a select group of technology and cybersecurity firms to assist in testing software vulnerabilities. Anthropic stated that due to the model’s immense power, if misused, it could cause severe consequences, necessitating restricted distribution.

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

Fable 5, publicly released this week, was described by Anthropic as a “Mythos-level model” equipped with additional safety measures enabling access to general users. The model had undergone review by both the U.S. government and the UK AI Safety Institute.

But following the disclosure of alleged security flaws, multiple government officials concluded the model must be removed immediately.

Saturday morning, David Sacks—former White House AI advisor and a staunch opponent of regulation—posted on X in support of the government’s export control action against Anthropic.

Sacks stated he did not view this “jailbreak” as trivial or inconsequential, nor did he see export controls as an attempt by the government to exert broader control over the AI industry. He wrote: “The government now wants Anthropic to fix the security issues, then lift the export controls and reintroduce Fable to public availability. They want this resolved quickly. Frankly, the government is confused: why is Anthropic, who has consistently said safety requests are their highest priority, now unwilling to comply?”

Sacks and other government officials had previously criticized Anthropic, accusing it of left-wing political bias and claiming the company incited panic by advocating 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 unprecedented levels. On March 3, citing Anthropic’s refusal to allow its AI tools to be used in mass domestic surveillance and autonomous weapons systems, the Pentagon classified the company as a supply chain risk.

Saturday, Sacks clarified that the existing tensions between the government and Anthropic were separate from the current export control decision. He wrote: “The government values Anthropic’s technological capabilities and believes this issue, while serious, should be easily solvable. Now the ball is in Anthropic’s court.”

[Original Link]

Original: BlockBeats