Trump Administration Backs Down on Anthropic AI Ban After Security Clash Over Fable 5 and Mythos 5

• Early June: Hype Meets a Sudden Halt
• Mid-June: National Security Panic and a Showdown in Washington
• Late June: Safeguards, Testing, and Limited Reopening
• June 30–July 1: Export Curbs Lifted and Global Return
• Competing Narratives: Safety, Power, and Politics

Trump Administration Backs Down on Anthropic AI Ban After Security Clash Over Fable 5 and Mythos 5
The abrupt global shutdown of Anthropic’s most advanced AI models has flipped into an equally sweeping restart, exposing a tense trial run of how the U.S. will police frontier AI while trying not to kneecap its own tech champions.

§Early June: Hype Meets a Sudden Halt

In early June, Anthropic rolled out Claude Fable 5, a consumer-facing system built on the same core technology as its more powerful Mythos 5 model, which had been limited to select organizations because of its cybersecurity capabilities. On June 12, the Commerce Department slapped both models with export controls, barring access by foreign nationals and forcing Anthropic to suspend Fable 5 and Mythos 5 for all users because it could not “verify nationality in real-time.”

§Mid-June: National Security Panic and a Showdown in Washington

Officials had been alarmed by a possible Fable 5 “jailbreak” and Mythos 5’s ability to find and exploit software vulnerabilities, treating the pair as potential national security risks. Anthropic executives rushed to Washington, arguing the order stemmed from a misunderstanding and highlighting that Mythos had already been confined to a narrow “Project Glasswing” program for vetted cybersecurity use. The clash followed earlier tensions, including a Pentagon move labeling Anthropic a supply-chain risk and an ongoing Anthropic lawsuit against the Trump administration.

§Late June: Safeguards, Testing, and Limited Reopening

To address jailbreak fears, Anthropic trained an improved safety classifier targeting the specific exploit and routing blocked Fable 5 queries to a less risky Opus 4.8 model. Axios reported that new safeguards now blocked the worried-about jailbreak 93% of the time, with the remaining 7% only surfacing previously known or patched flaws. Commerce’s Center for AI Standards and Innovation tested both old and new defenses, while the government allowed Mythos 5 to return for a select group of approved organizations on June 26.

§June 30–July 1: Export Curbs Lifted and Global Return

After weeks of negotiation, Commerce Secretary Howard Lutnick informed Anthropic it would “no longer need a license for exports or in-country transfers” of Fable and Mythos, citing “steps in close coordination with the US government to address the risks.” TechCrunch reported that Anthropic agreed to “proactively detect and address security risks,” collaborate on standards for future releases, and promptly notify officials of malicious use.

That same evening, the Trump administration formally lifted export controls on Fable 5, restoring public access after an 18-day suspension. Anthropic announced it would begin restoring Fable 5 and Mythos 5 access globally on July 1, including on major cloud platforms.

§Competing Narratives: Safety, Power, and Politics

Government officials framed the episode as a model for public–private cooperation and a win for U.S. AI leadership, with Lutnick saying his office had “worked closely with Anthropic to analyze and approve Fable 5 to ensure alignment across the US Government.” White House allies cast the outcome as an “America First” success that got “the best tech deployed as quickly and safely as possible.”

Industry voices and cybersecurity experts, however, questioned whether the initial export ban was more about leverage and political punishment than security, noting Anthropic had already pledged many of the required safeguards. With rival Asian labs racing toward Mythos-level systems, pressure mounted on Washington not to sideline a leading U.S. AI firm just as Anthropic moves toward an IPO.

Anthropic, for its part, has signaled it will expand red-teaming, 24/7 jailbreak monitoring, and pre-release testing with the government, even as Commerce reserves the right to reimpose curbs at any time. The Fable 5 saga now stands as an early template—and warning—for how future frontier models may be negotiated into existence.

Continue reading https://foxvector.com/stories/019f6191-ca66-3b25-734b-3c76c459879e