The U.S. government has issued an emergency export control directive ordering AI research firm Anthropic to immediately suspend access to its two most advanced models, Claude Fable 5 and Claude Mythos 5, for all foreign nationals. This action, taken just days after the models’ release, cites national security concerns related to a potential vulnerability allowing users to bypass the AI’s safety protocols, commonly known as “jailbreaking.” The broad scope of the directive forced Anthropic to disable access for its entire customer base to ensure compliance, impacting even its own international employees.
Key Takeaways
- The U.S. government mandated Anthropic halt access to its Fable 5 and Mythos 5 AI models for foreign nationals due to national security concerns over a jailbreak vulnerability.
- Anthropic complied with the order but contested the severity of the vulnerability, stating it is simple, already present in other industry models like GPT-5.5, and does not represent a significant security threat.
- The company warned that enforcing such a precedent could impede future advancements and deployments of cutting-edge AI models across the industry.
- A government advisor suggested Anthropic refused a request to fix the identified vulnerability, leading to the directive.
- Anthropic has previously clashed with the U.S. government over agreements related to AI safety and surveillance, highlighting ongoing tensions in AI governance.
The government’s directive did not detail the specific national security concern but is believed to stem from awareness of a method to circumvent Fable 5’s safeguards. Mythos 5, a more powerful model with fewer restrictions and enhanced capabilities for identifying cybersecurity exploits, was already accessible to a limited group of partners.
Anthropic has challenged the government’s assessment, characterizing the identified vulnerabilities as “relatively simple” and noting that similar capabilities are readily available in competing public models, including OpenAI’s GPT-5.5. The company indicated that the evidence provided by the government for the jailbreak was verbal and described a narrow, non-universal technique involving specific code analysis and flaw identification.
Despite complying, Anthropic expressed concern that this directive sets a “dangerous precedent.” The company warned that if similar standards were applied broadly, it could effectively halt the deployment of new AI models from all leading frontier AI providers. Access to Anthropic’s other AI models remains unaffected, and the company is working towards restoring access to Fable 5 and Mythos 5 as quickly as possible.
I’ve had a number of conversations with folks inside and outside government about the current situation with Anthropic, and here is what I believe to be true:
— As we know, Anthropic publicly released its Mythos class models earlier this week under the commercial name Fable.…
— David Sacks (@DavidSacks) June 13, 2026
Adding to the public discourse, David Sacks, co-chair of the President’s Council of Advisers on Science and Technology, shared his perspective on X. He alleged that a trusted partner of both Anthropic and the U.S. government discovered a jailbreak in Fable 5. According to Sacks, Anthropic CEO Dario Amodei reportedly refused a government request to either fix the vulnerability or disable the model, a stance Sacks contrasted with Anthropic’s public advocacy for AI safety and regulation.
Sacks further suggested that Anthropic prioritized the commercial release of its model over addressing the safety concern, leading to the government’s intervention. He indicated that the administration views the issue as resolvable and hopes Anthropic will address the vulnerability to allow Fable 5’s public release to resume. Sacks stated, “The ball is in Anthropic’s court.”
This incident follows a previous disagreement earlier this year where Anthropic declined to sign expanded agreements that would have permitted mass domestic surveillance and the use of fully autonomous lethal weapons systems. This led to criticism from former President Donald Trump and a “supply chain risk” designation from the Department of Defense, which Anthropic is contesting in court. Recent reports suggest a cooling of tensions as the government has sought access to other Anthropic models.
Long-Term Technological Impact: Balancing Innovation and Control
This governmental action against Anthropic raises significant questions about the future trajectory of AI development and deployment, particularly concerning the interplay between rapid innovation and national security imperatives. The imposition of an emergency export control directive, even if disputed by the AI firm, signals a growing willingness by governments to intervene directly in the commercial rollout of advanced AI. This could have profound implications for the blockchain and Web3 space, where decentralized infrastructure and AI integration are increasingly becoming cornerstones of new applications and platforms. While blockchain technology itself offers enhanced transparency and security, the centralized nature of AI model development and control, as highlighted by this case, presents a potential bottleneck.
The concern that such directives could “halt all new model deployments” suggests a risk of stifling innovation. For Layer 2 scaling solutions and other blockchain advancements that rely on sophisticated AI for optimization, security, and user experience, regulatory uncertainty or overly restrictive oversight could slow progress. The industry may need to explore more decentralized AI development models or robust self-regulatory frameworks to preemptively address security concerns without impeding the pace of technological evolution. The debate also underscores the need for clearer, industry-wide standards for AI safety and security, developed collaboratively by researchers, developers, and policymakers, to foster responsible innovation in AI and its integration with emerging technologies like blockchain.
Original article : decrypt.co