Anthropic said late Tuesday that it would restore global access to its most advanced language models, Claude Mythos 5 and Claude Fable 5, on Wednesday after the U.S. Department of Commerce removed export‑control restrictions. The announcement came after a two‑week engagement with federal officials, during which the company demonstrated the models’ cybersecurity safeguards and agreed to a framework for assessing potential jailbreak vulnerabilities.
"Although we have reached a constructive resolution, these events have made clear that the industry needs a consistent way to assess and fix potential 'jailbreaks' of AI models," Anthropic wrote in a statement. "A shared standard… would help AI developers triage new findings as they arise, launch highly capable models with greater safety, and communicate the level of risk consistently to government and industry partners."
Government clearance and rollout plan
The Department of Commerce confirmed it had lifted the export controls that forced Anthropic to block the models for all foreign nationals in early June. With the clearance in place, Anthropic will begin restoring access across its cloud partners—Amazon Web Services, Microsoft Foundry and Google Cloud—"as quickly as possible," the company said. Access will be phased throughout the day, and users can already tap the newly released Sonnet 5 model.
Commerce Secretary Howard Lutnick, who oversees the regulatory Center for AI Standards and Innovation, posted on X that the agency had worked closely with Anthropic to analyze and approve Fable 5, reinforcing the United States' leadership in AI safety.
For paying customers, Anthropic set a temporary usage ceiling: Pro, Max, Team and enterprise plans may consume up to 50 % of their weekly token limits on Fable 5 through July 7. After that date, the model will require additional usage credits. Mythos 5 will initially be available only to a select group of U.S. companies participating in Project Glasswing.
The rollout follows a broader shift in federal AI policy. An executive order signed on June 2 mandates that the Defense Department coordinate with AI firms to obtain government review before releasing frontier models. While Mythos 5 and Fable 5 were blocked under a separate legal authority, the order signals a more hands‑on approach to AI governance.
Anthropic, which launched Fable 5 in June as the first publicly available Mythos‑level model, had to pull the model after a week when the export‑control order required a blanket block for any foreign user. The company complied by removing the model entirely, citing the need to meet the regulator’s demand.
Industry observers note that Anthropic’s push for a universal evaluation framework could set a precedent for future AI releases. By proposing a shared standard, the firm hopes to streamline the triage of vulnerabilities and provide clearer risk communication to both regulators and customers.
Anthropic’s CEO and senior leadership have not disclosed the exact timeline for expanding Mythos 5 beyond the initial U.S. cohort. However, the company emphasized its commitment to “launch highly capable models with greater safety” and to work with partners to refine the proposed standards.
Customers who rely on Anthropic’s models for enterprise workflows, content generation, or research will likely see a brief interruption before services normalize. The company thanked users for their patience and pledged to keep them updated as access fully restores.
Cet article a été rédigé avec l'assistance de l'IA.
News Factory APP - actualités agentiques pour booster votre SEO et AEO.