Washington Backs Down on Anthropic Restrictions, Easing Rules on Mythos 5 and Fable 5 AI Models

A Partial Reprieve for Anthropic’s Cybersecurity Engine

The United States government has partially eased its recent restrictions on Anthropic, allowing the artificial intelligence safety startup to redeploy its highly powerful cybersecurity model, Claude Mythos 5. In a letter sent to the company, Commerce Secretary Howard Lutnick confirmed that selected partners would be permitted to access the model once again. Lutnick noted that he determined appropriate safeguards were in place to allow certain trusted partners to use Claude Mythos 5, adding that Anthropic has committed to collaborate with the federal government on future protocols, standards, and releases. This decision followed two weeks of intense, daily negotiations between the tech firm and federal officials.

Under the newly relaxed guidelines, approximately 100 trusted organizations will regain access to the cybersecurity system. These entities include major American corporations, government agencies, and organizations responsible for operating and defending critical domestic infrastructure. While this move restores access for a select group of high-priority users, Anthropic is continuing its discussions with Washington to expand access to the model further.

The Security Concerns That Sparked the June 12 Shutdown

The partial rollback of these restrictions marks a significant de-escalation in a sudden clash between the Trump administration and the artificial intelligence developer. On June 12, 2026, the government issued sweeping export controls that forced Anthropic to completely cut off access to both Claude Mythos 5 and its consumer-oriented counterpart, Fable 5. The federal directive was so strict that it even barred foreign nationals employed by Anthropic within the United States from accessing the company's own systems.

This drastic intervention was prompted by warnings from rival tech firms, including Amazon, which cautioned officials that the advanced models could be jailbroken and exploited by malicious actors. Furthermore, federal authorities expressed deep concern over intelligence indicating that a hacking group linked to China may have successfully accessed Mythos 5. These dual anxieties about foreign espionage and potential cyberweaponry led to the rapid implementation of the mid-June blockade.

Fable 5 Set for General Public Release

In addition to the Mythos 5 redeployment, the Trump administration is poised to lift export limits on Fable 5 as early as Tuesday night. Fable 5 serves as the consumer-friendly version of Mythos 5, designed to bring advanced vulnerability-detection capabilities to the public while incorporating strict guardrails to prevent users from deploying the software to launch cyberattacks of their own. The initial ban on Fable 5 had caused widespread frustration across the domestic tech sector and among European allies, who were eager to utilize Anthropic's tools to identify and repair security flaws in their networks before foreign adversaries could exploit them.

The expected policy shift highlights the delicate balance the White House is trying to maintain between mitigating national security threats and fostering rapid technological innovation to keep pace with Chinese competitors. This regulatory compromise comes on the heels of similar moves in the industry, including an agreement by Anthropic's primary rival, OpenAI, to limit the availability of its own frontier models. To prevent future regulatory standoffs, Anthropic and administration officials have discussed establishing a standardized technical assessment framework to evaluate the risks of upcoming models before they are released.

As Washington establishes these selective access pipelines and technical assessments, the line between safeguarding national infrastructure and choking commercial AI development remains dangerously thin.