The White House is said to have factored a possible China‑linked breach into its decision to impose export controls on Anthropic's flagship AI systems, Mythos and its companion model Fable. According to a report from Semafor, officials suspect a group tied to the Chinese government accessed the models, raising alarms about national‑security implications.
Anthropic's Mythos, touted as one of the most powerful generative AI models, is currently subject to strict export restrictions aimed at limiting its spread to foreign actors. If Chinese operatives obtained versions of Mythos 5 or Fable 5, Washington argues the United States could face a "serious national security risk." The concern centers on the potential for reverse engineering through a process known as distillation, where a less‑capable "student" model learns to mimic the behavior of a more advanced system.
White House officials have not publicly confirmed the allegations. A post on the social platform X by Trump adviser David Sacks highlighted recent reports that Anthropic's models could be "jailbroken," a claim the company has denied, but Sacks made no mention of China. Anthropic declined to comment on the specific report, though a spokesperson told Semafor that China was not raised during the administration's export‑control discussions.
Should the breach be verified, it would not be Anthropic's first encounter with unauthorized access. The company previously disclosed that a Discord group gained entry to Mythos for roughly two weeks before the breach was detected and the access was revoked. That episode underscored the challenges of safeguarding cutting‑edge AI systems against determined actors.
Experts note that once a model is in the hands of a foreign entity, even limited exposure can enable the creation of derivative versions that retain much of the original's capability. Distillation, in particular, can produce a compact model that mimics the performance of its larger predecessor, potentially allowing adversaries to sidestep export restrictions.
The administration's move reflects a broader U.S. strategy to tighten oversight of advanced AI technologies. Recent policy discussions have centered on balancing innovation with security, especially as AI systems become increasingly integral to defense, intelligence, and critical infrastructure.
While the White House has not released an official statement confirming the China link, the report adds pressure on policymakers to consider more aggressive measures. Industry observers will watch how Anthropic and other AI developers respond to heightened scrutiny and whether additional safeguards become standard practice.
Este artigo foi escrito com a assistência de IA.
News Factory APP - notícias agênticas para impulsionar seu SEO e AEO.