Anthropic Adds ID Checks as Fable 5 Remains Banned

Anthropic updated its policy language earlier in June to require users to verify age or identity "in certain circumstances" by uploading an image of a government-issued ID, per reporting by Gizmodo. The company announced on June 12 in a blog post titled "Statement on the US government directive to suspend access to Fable 5 and Mythos 5" that the U.S. government issued an export control directive to suspend access to Fable 5 and Mythos 5 for "any foreign national, whether inside or outside the United States," and that Anthropic received the directive at 5:21pm (ET). The blog post says the directive forced Anthropic to "abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance," and that the government showed a demonstration of a method to bypass, or "jailbreak," safeguards that could be used to identify software vulnerabilities. Reporting by The Verge, The Guardian, Forbes, NBC News, and Al Jazeera corroborates the export-control order and the company's response.

Industry reporting places the new ID/age-verification requirement in the context of Anthropic's use of third-party verification tooling. Gizmodo reports Anthropic tapped Peter Thiel-backed Persona Identities to build the verification mechanism, and cites researchers who found Persona had been checking users' biometric data against government watchlists, a finding that previously led Discord to abandon a partnership plan. Persona's known partners, as reported in public coverage, include OpenAI, Lyft, Square, Reddit, and LinkedIn.

Export-control enforcement moving from hardware and chips toward restricting access to deployed models is an emerging policy pattern. Industry reporting frames the U.S. directive as part of a broader escalation in efforts to limit foreign access to advanced models. Public coverage records a dispute between Anthropic and U.S. officials over the scale of the vulnerability in question; Anthropic's blog says testers found only "relatively simple" vulnerabilities and that no tester has found a "universal jailbreak." The company blog describes its safeguards as "substantially more effective than those of any previously deployed model," while noting the industry-wide challenge that perfect jailbreak resistance may not be attainable.

Developers and deployers of frontier models should treat this episode as a signal that operational controls and identity solutions are necessary but not sufficient to address state-level export-control actions. The policy lever at play is not limited to technical mitigations - it hinges on government assessments of risk and demonstrated bypasses. For teams building globally available APIs or models, the combination of enhanced user verification and regulatory constraints increases the complexity of access control, auditability, and legal compliance.

• •Whether the U.S. Commerce Department or other agencies publish technical details or a formal rationale for the directive, which would affect how vendors and auditors assess risk.
• •Any follow-up from Anthropic showing changes to Fable 5 safeguards or new verification flows beyond the Persona integration, and whether those changes are accompanied by third-party assurance testing.
• •Broader industry and government responses, including whether other model providers face similar export-control scrutiny or are asked to adopt identity-based access controls.

The addition of ID checks intersects with privacy and adoption trade-offs seen in other age-verification efforts. Public reporting shows those debates center on whether identity gating meaningfully alters the underlying export-control calculus, or simply introduces new privacy and operational burdens without resolving the channel by which a government judges a model to be a risk.