Two weeks after a federal ban halted Anthropic’s rollout of its most advanced cybersecurity models, the Trump administration has softened its stance. In a letter seen by Semafor, Commerce Secretary Howard Lutnick told Anthropic’s chief compute officer Tom Brown that the government will allow the Claude Mythos 5 model to be redeployed to more than 100 specific U.S. government agencies and private firms that protect critical infrastructure.
The approval marks a notable shift. The original ban, imposed on June 12, barred both Mythos 5 and its sibling model Fable 5 after security researchers demonstrated that the models’ guardrails could be bypassed. At the time, the order also prohibited non‑American employees—including Anthropic’s own foreign staff—from accessing the models.
Lutnick’s missive says the administration has determined that “appropriate safeguards are in place” for the selected partners. Those safeguards now cover non‑American personnel at the approved organizations, a reversal of the earlier restriction. The list of eligible entities includes a mix of federal agencies, state‑level bodies, and private companies that operate or defend critical infrastructure, though the exact names were not disclosed.
Anthropic’s response on X confirmed the development. The company said it has been working closely with the U.S. government since the ban’s inception and that the notification allows Mythos 5 to be restored quickly for the designated partners. Anthropic added that it continues to negotiate broader access for Mythos 5 and hopes to make Fable 5 available for general use again.
While Mythos 5 receives a green light, the administration’s directive does not mention Fable 5. The latter model, released a few days before the ban because of its purportedly stronger protections, remains out of reach for all but Anthropic’s internal teams.
Industry observers note that the move reflects a balancing act between national security concerns and the need for advanced AI tools in protecting the nation’s digital arteries. Cybersecurity experts have long warned that sophisticated language models can both aid defenders and be weaponized by adversaries. By limiting access to a vetted group, the government aims to harness the technology’s defensive potential while mitigating misuse.
Anthropic has not commented further on the specifics of the safeguards or the timeline for expanding access to Fable 5. The company’s statement emphasized a “rapid” restoration for Mythos 5 and an ongoing dialogue with federal officials.
The episode underscores the growing tension between AI innovation and regulatory oversight. As more powerful models emerge, policymakers are forced to draw lines that protect public safety without stifling the benefits these tools can deliver to critical sectors.
Dieser Artikel wurde mit Unterstützung von KI verfasst.
News Factory APP - agentische News für besseres SEO & AEO.