The US government has reportedly directed Anthropic to suspend access to its Fable 5 and Mythos 5 artificial intelligence models for all foreign nationals, including the company’s own foreign national employees. Following the directive, Anthropic has temporarily disabled access to these models for all customers worldwide.
In a statement, Anthropic said the order was issued under national security authorities and specifically applies to foreign nationals “whether inside or outside the United States.” The company clarified that access to its other AI models remains unaffected.
“The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance,” Anthropic said.
The move marks one of the most significant government interventions yet in the deployment of advanced AI systems, coming just days after Anthropic unveiled its new Mythos-class models. Fable 5 was designed as the publicly accessible version, while Mythos 5 was placed under stricter controls.
According to Anthropic, it received the directive on June 12 at 5:21 pm ET, although the government did not provide detailed information regarding the national security concerns behind the decision.
The company said it believes the action is linked to a potential method of bypassing safeguards in Fable 5, commonly referred to as “jailbreaking.” Anthropic stated that it reviewed a demonstration of the technique, which reportedly exposed only a small number of previously known and minor vulnerabilities.
It further argued that such weaknesses are relatively simple and can also be identified using other widely available AI models. Anthropic maintained that Fable 5 had undergone extensive safety testing, including collaboration with US and UK AI safety institutes, third-party evaluators, and internal red-teaming before release.
“These tests showed that Fable’s safeguards are substantially more effective than those of any previously deployed model,” the company said.
Anthropic added that no tester had identified a “universal jailbreak” capable of broadly bypassing safeguards or enabling unrestricted cyber capabilities. The company also noted that achieving complete jailbreak resistance is currently not feasible for any AI system.
It described its approach as “defense in depth,” designed to limit jailbreak risks by making them difficult or costly to execute, while also combining monitoring systems to detect and mitigate misuse. As part of this strategy, Anthropic requires a 30-day data retention policy for Fable users to enhance security oversight.
The company disagreed with the government’s assessment, stating that it had not been presented with evidence of any serious or harmful jailbreak. According to Anthropic, the concern was based on a narrow capability involving tasks such as identifying and fixing software vulnerabilities in codebases—functions it claims are already available in other AI models, including OpenAI’s GPT-5.5.
Anthropic said it is complying with the legal directive but does not agree with the decision, warning that such standards could severely restrict future AI deployments if applied broadly across the industry.
While the company supports regulatory oversight of AI safety, it argued that such actions should follow a transparent and structured legal process grounded in technical evidence.
The suspension is expected to impact customers who had already integrated Fable 5 and Mythos 5 into workflows across software development, research, and enterprise automation, though Anthropic said efforts are underway to restore access where possible.
Recent Random Post:
