Anthropic Disables Fable and Mythos After Export Order

Anthropic disabled access to its most advanced models, Fable 5 and Mythos 5, after receiving a U.S. export-control directive that barred access by any foreign national, Reuters reports. CNBC reports the company received the directive at 5:21 p.m. ET on Friday and said it suspended access for all customers to ensure compliance. Fortune and Anthropic's public statement say the government letter "did not provide specific details" about the underlying national security concern. Anthropic apologized to customers for the disruption in a post on its site, writing, "We apologize for this disruption to our customers," per Fortune and Anthropic's statement.

Public reporting attributes the government's action to a perceived method to "jailbreak" safeguards that would allow Fable 5 to surface advanced cybersecurity capabilities tied to Mythos 5, Reuters and Fortune report. Sources describe the cited jailbreak as narrow rather than universal; Anthropic told reporters it believed the same technique could be used against other high-end publicly available models, including OpenAI's GPT-5.5, Fortune reports. For practitioners, narrow jailbreaks that expose specific capabilities present a different mitigation problem than universal bypasses because they often depend on particular prompt patterns or environment conditions.

The U.S. Commerce Department used export controls to block foreign access to commercial AI models, according to Fortune and Axios reporting. Public coverage frames the decision as part of a broader trend of tighter government scrutiny over advanced AI capabilities and their cross-border distribution, with this action notable because it forced a commercial model pull after a same-day directive. Observers in technology and policy coverage described the decision as an escalation in how national-security rules are being applied to AI model exports (Politico, Axios, Time).

For AI teams and platform operators, the incident highlights operational risk from regulatory orders that can require immediate access controls across global user bases. Deployers that serve multinational customers or employ non-U.S. staff may face abrupt availability constraints if similar export controls are applied, increasing the value of geo- and identity-aware access controls, compartmentalized model rollouts, and legal-compliance workflows. Several outlets reported that Anthropic disabled the models for all users because the directive covered foreign nationals even when they were inside the United States (CNBC, Fortune).

Observers should track:

• •whether the Commerce Department or other agencies publish further details justifying the national-security finding
• •whether regulators apply the same export-control standard to other firms or models
• •how cloud and enterprise customers respond contractually and operationally. Reuters and WSJ reporting indicate Anthropic engaged with U.S. officials following the order; Politico and WSJ describe rapid follow-up activity in Washington but did not publish any government technical evidence publicly. Longer term, practitioners will watch whether this enforcement approach leads to standardized government disclosure of exploit techniques, a tighter industry standard for model hardening, or narrower, time-limited restrictions