AI Read the original on Decrypt 2 min read 0

US government orders Anthropic to suspend Claude Fable and Mythos

The United States government has issued an emergency export control directive requiring Anthropic to immediately suspend access to its most advanced artificial intelligence models, Claude Fable 5 and Claude Mythos 5. The order stems from national security concerns regarding a potential jailbreak vulnerability that could allow unauthorized users to bypass safety protocols. While Anthropic complied with the mandate by disabling the models for all customers, the company maintains that the government's assessment of the risk is disproportionate.

Дональд Трамп виступає за трибуною з печаткою президента США на фоні американських прапорів та графічних елементів у червоно-синіх тонах.
Дональд Трамп виступає за трибуною з печаткою президента США на фоні американських прапорів та графічних елементів у червоно-синіх тонах. · Image source: Decrypt

According to Decrypt, the U.S. government has mandated a total suspension of access to Anthropic's high-end AI models, Claude Fable 5 and Claude Mythos 5. The directive applies to all foreign nationals, including the company's own staff, regardless of whether they are located inside or outside of the United States. Because of the sweeping scope of this order, Anthropic was forced to take the models offline for its entire global customer base to ensure full regulatory compliance.

National security and jailbreak concerns

The government's intervention follows reports that a method exists to "jailbreak" the publicly available Fable 5 model. While the specific details of the national security threat were not disclosed in the official letter, officials believe the vulnerability poses a significant risk. The Mythos 5 model, which features fewer guardrails and possesses specialized capabilities for identifying cybersecurity exploits, was originally restricted to select partners before this intervention.

Anthropic has publicly disputed the severity of these findings. The company stated that it reviewed a demonstration of the alleged jailbreak technique and concluded that the vulnerabilities are relatively simple. Furthermore, Anthropic argues that other commercially available models can perform similar tasks without any specific bypass techniques required. The company noted that:

  • The government has only provided verbal evidence of a narrow, non-universal jailbreak.
  • The exploit essentially involves asking the model to read a codebase and identify software flaws.
  • Competing models, such as OpenAI's GPT-5.5, already possess these capabilities publicly.
  • Industry implications and safety debates

    Despite complying with the order, Anthropic warned that this move establishes a dangerous precedent for the technology sector. The company expressed concern that if such standards are applied universally, it could effectively halt all new deployments of frontier AI models across the industry. While access to other Anthropic products remains unaffected, the company is currently working to restore access to the suspended models as quickly as possible.

    The situation has sparked internal friction regarding corporate ethics and safety branding. David Sacks, co-chair of the President’s Council of Advisers on Science and Technology, suggested that Anthropic's refusal to immediately fix the reported jailbreak led to the government's reluctant export control action. Sacks claimed that "Anthropic prioritized the continued offering of the consumer model over safety," noting a contradiction between the company's public ethos and its actions in this instance. The administration remains hopeful that Anthropic will resolve the technical flaw so that Fable 5 can eventually return to public release.

    FAQ

    Why were the Claude Fable 5 and Mythos 5 models suspended?
    The US government issued an emergency export control directive due to national security concerns. Officials believe a jailbreak vulnerability exists that could allow unauthorized users to bypass safety protocols, specifically regarding identifying cybersecurity exploits and software flaws.
    What is Anthropic's response to the government order?
    Anthropic complied with the mandate but disputes the severity of the risk. The company argues the vulnerabilities are simple, that other models like GPT-5.5 already possess similar capabilities, and that the move sets a dangerous precedent for the AI industry.
    Telegram

    Fresh news on our Telegram

    Get instant alerts for new posts in «AI»

    @proaiandevenmore