Government Mandates Halt to Anthropic's Flagship AI Models
Anthropic, a leading artificial intelligence startup, has been compelled by the Trump administration to disable its most advanced AI models, Fable 5 and Mythos 5, for all users globally. This drastic measure follows an export control order issued by the White House on Friday, citing national security concerns. The directive specifically mandated suspending access for any foreign national, including Anthropic's own foreign employees.
The company stated that to ensure compliance with the order, it had to "abruptly disable" access to the models for all customers. This action marks an unprecedented intervention by the U.S. government into the deployment of frontier AI models developed by an American company.
National Security Concerns and Alleged "Jailbreaks"
The Trump administration's decision was reportedly driven by fears of "jailbreaks" or methods to bypass the safeguards built into Fable 5 and Mythos 5. Administration officials claim that Anthropic's "recklessness" in addressing these vulnerabilities led to the export controls. Sources indicate that Amazon and five other companies notified the White House of their ability to perform these jailbreaks, with a message from Amazon's CEO, Andy Jassy, reportedly being particularly influential.
Furthermore, a report from Semafor suggests that the White House was also concerned that a "China-linked group" had gained access to a Mythos-class model, despite Anthropic's policy of not allowing access from China. Anthropic, however, has disputed the severity of the identified vulnerabilities, characterizing them as "previously known, minor vulnerabilities" and arguing that other publicly available models could also discover them without requiring a bypass. The company maintains that its safeguards are "substantially more effective than those of any previously deployed model."
High-Stakes Meetings in Washington D.C.
In response to the government's order, senior technical staff from Anthropic are currently in Washington D.C., engaging in both virtual and in-person meetings with White House officials. The company is reportedly "mobilizing quickly to make amends" with the Trump administration. These discussions aim to resolve the dispute and potentially restore access to the powerful AI models.
The situation highlights growing regulatory risks for leading AI developers, as governments increasingly grapple with the security implications of advanced AI. This dispute also comes at a critical time for Anthropic, which has confidentially filed for a U.S. initial public offering. Some observers view the government's action as a "corrupt political shakedown," suggesting it may be an attempt to secure a "massive 'donation'" or equity in Anthropic's upcoming IPO.
Escalating Tensions and Broader Implications
This latest conflict reignites a long-standing feud between Anthropic and the Trump administration. Earlier this year, the company clashed with the Pentagon over the use of its technology for military and surveillance purposes, leading to Anthropic being designated a U.S. supply-chain risk. While a federal judge issued a temporary injunction against those actions, the current export controls represent a significant escalation.
The administration's stance suggests a "government messaging 101" approach: if an AI company, particularly one advocating for AI safety, appears reluctant to address security concerns, the government will take a tough stance. Anthropic, however, has expressed its disagreement with the government's rationale, stating that a "narrow potential jailbreak" should not warrant recalling a commercial model deployed to hundreds of millions of people. The company warns that applying such a standard across the AI industry could significantly slow the deployment of new frontier models.