Semafor: China-Linked Group Suspected of Accessing Anthropic’s Top-Secret AI – Bitcoin News

What happened

Anthropic close global access to two AI models, Mythos and Fable 5, following a Commerce Department directive citing national security concerns. Since verifying user nationality per real-time API session is not technically feasible, Anthropic completely removed both models from the market rather than attempt selective enforcement.

Semafor, in a exclusive report by Reed Albergotti published this week, said the export control decision was partly motivated by suspicions that a China-linked group had accessed Mythos. If the Chinese government obtained the model, U.S. officials feared that Beijing could reverse engineer it through a process known as “distillation,” effectively replicating its capabilities.

It remains unclear which specific organization was involved, how access occurred, or how the White House became aware of it.

What is the myth

Anthropic launched Claude Mythe Overview in April 2026 as a frontier AI model built around autonomous cybersecurity offense and defense. The company deemed it too dangerous for public release, citing its ability to identify and exploit previously unknown software vulnerabilities on a large scale.

Key capabilities documented by Anthropic include:

• Autonomously identified thousands of high-severity zero-day vulnerabilities across major operating systems and browsers
• Discovery of 27-year-old remote crash vulnerability in OpenBSD, used in critical infrastructure
• Chained Linux kernel exploits allowing full privilege escalation
• 83.1% accuracy on the CyberGym vulnerability reproduction benchmark

AI-accelerated attacks could reduce what currently takes months of expert work to minutes, according to Anthropic. Cybercrime already costs around $500 billion per year worldwide.

Access was limited to approximately 12 named partner companies and more than 40 additional organizations through Project Glasswing, a defensive initiative designed to find and fix vulnerabilities before bad actors can exploit them.

Fable 5 jailbreak

A second model, Fable 5, a consumer version of Mythos built with safeguards against advanced cyber use, was launched in early June. White House AI advisor David Sacks posted on X that Anthropic CEO Dario Amodei had been informed that Fable 5 had been jailbroken but would have minimized the risk and refused to remedy it.

Sacks said the administration “appreciates Anthropic’s technical capabilities” and described the situation as “serious but easily resolved,” adding that “the ball is in Anthropic’s court.”

Anthropic disputed the seriousness of the jailbreak, saying it was limited in scope and involved problems already present in other publicly available models. The company also said the White House did not raise Chinese access concerns in direct discussions, which focused on Fable 5 and the circumvention technique.

For what Cryptocurrency The platforms are watching

Mythos has a direct relationship with the security of digital assets. Anthropic’s work using the model has identified millions of potential blockchain exploit exposure and model can probe autonomously smart contracts, decentralized finance ( Challenge), APIs and weak retention systems.

Major exchanges, including Coinbase, reportedly initiated discussions on defensive access via Project Glasswing. Suspending global access terminates these agreements for unapproved parties and removes an important defensive tool for operators not on the list of approved partners. More recently, just before the ban of Claude Fable 5 and Mythos 5, the founder of Zcash claimed Mythos audited the entire Zcash chain and found no bugs.

Security professionals who relied on the model’s vulnerability scanning capabilities now have no path forward. This gap benefits any actor who has already obtained a copy.

AI competition between the United States and China

The incident is part of a larger pattern. Semafor’s reports and previous disclosures link Chinese state-linked actors to previous attempts to use anthropogenic models for cyberespionage. A campaign documented by Anthropic, labeled GTG-1002has linked Chinese state actors to AI-assisted attacks against around 30 entities, including financial institutions.

German cybersecurity officials have separately warned that Chinese AI vendors appear to be cutting back on public updates while developing capabilities in-house equivalent to those of Mythos. China has the data center infrastructure and computing capacity to continue on this development path.

The suspension of access to Mythos and Fable 5 remains in effect as of June 14, 2026. Anthropic has not announced a timeline for restoring service.