The White House effectively terminated public access to Anthropic’s Fable 5 artificial intelligence model on Friday night, following an urgent intervention by Amazon and other industry stakeholders regarding national security vulnerabilities. The model, a general-use version of Anthropic’s Mythos AI, was pulled from operation after the administration issued a letter invoking sweeping export control regulations, according to reports confirmed by multiple industry sources.
The administrative action, which occurred on the evening of June 14, represents a significant escalation in how the U.S. government monitors the deployment of advanced AI. While Anthropic had informed federal officials of the planned June 9 release, the subsequent security concerns raised by Amazon prompted a rapid, high-level scramble within the executive branch to restrict the technology’s availability.
The Origins of the Security Dispute
The decision to restrict access followed a disclosure from Amazon, a primary investor in Anthropic, regarding the potential for “jailbreaking” the underlying Mythos model. According to industry sources, Amazon officials contacted the White House on Thursday evening to present evidence that the model could be manipulated to bypass safety protocols, potentially exposing sensitive data or creating national security risks.
An Amazon spokesperson confirmed the company’s role in advising the government on potential risks, stating, “As a leading cloud provider that serves a large number of private and public sector customers, it’s not uncommon for governments to seek our counsel on potential security risks. When they occur, we don’t share the details of these discussions.”
Anthropic, however, has maintained that the risks identified were overstated. Sources close to the company indicated that CEO Dario Amodei and his team held discussions with the administration throughout Friday, arguing that the reported vulnerabilities were relatively simple and could be replicated across various AI models. They contended that the findings did not reflect a fundamental flaw in the safety architecture of Fable 5.
Regulatory Impact and Export Controls
The White House formalized the takedown through a letter sent to Anthropic at approximately 5:20 p.m. ET on Friday. By 10:00 p.m. ET, public access to the Fable model had been suspended. This action utilized export control mechanisms that effectively categorize the models as restricted technology, limiting access not only for foreign adversaries but also for foreign nationals within the United States.
The implications of this policy shift are immediate for the technology sector. By applying these controls to a specific company’s release, the administration has signaled a move toward a de-facto licensing regime for high-level AI models. Industry analysts suggest that this creates a new precedent where companies may be required to seek implicit or explicit government approval before deploying models that surpass established safety benchmarks.
Luta Security CEO Katie Moussouris, who reviewed the research report provided by Amazon, challenged the severity of the government’s reaction. “The government’s response seems way out of line with what’s actually in the research report,” Moussouris stated. She argued that the vulnerabilities identified were a product of researchers asking questions typical of defensive security testing, which is essential for hardening AI systems against future attacks.
Industry Consequences and Future Oversight
The incident highlights a growing tension between rapid AI innovation and the federal government’s mandate to protect critical infrastructure. Administration officials have stated that they do not view all AI models as threats, but emphasized that any model reaching or exceeding the performance threshold set by the Mythos architecture will face heightened scrutiny. Future deployments of this caliber will likely require advanced coordination with the government’s national security apparatus.
Within Anthropic, the sudden restriction has created operational challenges, particularly regarding the ability of foreign-born employees to engage with the company’s most advanced tools. The company’s leadership remains in dialogue with the administration to clarify the scope of the export controls and determine the path forward for their research and development cycles.
As the sector awaits further guidance on the specific criteria for these new export requirements, the episode serves as a clear indicator of the White House’s intent to maintain a “hardened” stance on AI security. The government has signaled that it will not hesitate to use its regulatory authority to pause releases if it determines that a company has been “overly confident” in its safety assessments.
The administration has not yet announced a formal date for the next public briefing regarding the specific export control standards for general-purpose AI. Updates on federal AI policy and potential adjustments to the current restrictions are expected to be published through official White House and Department of Commerce channels as the interagency review process continues.