The Trump administration announced Wednesday that it is lifting export controls on Anthropic’s flagship AI models, Mythos 5 and Fable 5. In a letter from Commerce Secretary Howard Lutnick to Anthropic co‑founder Tom Brown, the department declared that a license is no longer required for the export, re‑export, or in‑country transfer of either model, including deemed exports.
The move comes after weeks of negotiations between Anthropic and the Commerce Department, sparked by a June order that forced the company to suspend foreign‑national access to its most powerful systems. The original restrictions limited the models to a handful of U.S. companies and government agencies, citing national‑security concerns over potential misuse.
Anthropic agreed to a series of concessions designed to ease those concerns. The company pledged to proactively detect and remediate security risks associated with the models, and to work closely with the U.S. government on protocols, standards, and future releases. Those commitments were outlined in Lutnick’s letter, which emphasized Anthropic’s “diligent” cooperation on safeguards for Mythos, Fable, and any subsequent models.
Strengthening AI Safety Measures
Key to the agreement is Anthropic’s promise to reduce the number of “jailbreaks” – attempts by users to bypass built‑in safety restrictions and unlock higher‑risk capabilities. Officials have long worried that the Mythos model, in particular, could be weaponized for cyber‑attacks if its safeguards were compromised. By agreeing to enhance detection mechanisms and share threat intelligence with the Commerce Department, Anthropic hopes to prove that the risk of uncontrolled exploitation is manageable.
The administration’s national cyber director, Sean Cairncross, has been instrumental in shaping the deal, working alongside Lutnick to ensure that the company’s safety roadmap aligns with broader U.S. security objectives. While Anthropic previously argued that eliminating jailbreaks entirely was impossible, the new stance reflects a pragmatic shift: the firm will focus on building more robust safeguards and demonstrating measurable reductions in exploit attempts.
Industry observers note that the relaxed export regime could accelerate the commercial rollout of the models, opening doors for foreign partners that were previously barred. Anthropic’s CEO Dario Amodei, who recently stepped back from direct meetings with officials, has left Tom Brown to handle the dialogue. Sources say Brown’s rapport with the administration helped smooth the path to the agreement.
Critics, however, warn that easing restrictions may still leave gaps in oversight. They argue that even with stronger safeguards, the sheer power of Mythos and Fable could enable sophisticated attacks if adversaries discover new vulnerabilities. The administration counters that the collaborative monitoring framework established with Anthropic offers a realistic balance between innovation and security.
For now, the Commerce Department’s decision signals a willingness to work with AI developers rather than impose blanket bans. The letter, viewed by WIRED, marks the latest chapter in a broader debate over how the United States should regulate cutting‑edge AI technologies while preserving their economic and scientific benefits.
Anthropic’s next steps include publishing detailed safety reports, rolling out updated detection tools, and engaging with international partners under the new, unrestricted export terms. The company’s compliance timeline has not been disclosed, but officials expect measurable progress within the coming months.
Questo articolo è stato scritto con l'assistenza dell'IA.
News Factory APP - notizie agentiche per potenziare il tuo SEO e AEO.