Anthropic shut down worldwide access to two of its AI models on June 13, 2026. The move came after a Commerce Department directive citing national security concerns.
NEW: China-linked access to Anthropic's Mythos AI possibly sparked the White House CRACKDOWN that ultimately led to export restrictions, per The Verge. pic.twitter.com/rhgD28oRoV
— Coin Bureau (@coinbureau) June 15, 2026
The two models pulled were Mythos and Fable 5. Anthropic said it could not verify user nationality in real time per API session, so it removed both models entirely rather than attempt partial enforcement.
Semafor reported that the decision was partly driven by suspicions that a China-linked group had accessed Mythos. The report was based on an exclusive by journalist Reed Albergotti.
U.S. officials feared that if China obtained the model, it could replicate its capabilities through a process known as “distillation.” This involves training a separate AI on outputs from the advanced model.
Anthropic said the White House did not raise Chinese access concerns during their direct discussions. Those talks focused on Fable 5 and a reported jailbreak of that model.
Mythos was launched in April 2026 as a frontier AI built for cybersecurity. Anthropic considered it too dangerous for public release.
The model can autonomously find and exploit unknown software vulnerabilities. Anthropic documented that it identified thousands of high-severity zero-day flaws across major operating systems and browsers.
It found a 27-year-old remote crash vulnerability in OpenBSD and achieved 83.1% accuracy on the CyberGym vulnerability benchmark. Anthropic estimated it could compress months of expert hacking work down to minutes.
Access had been limited to around 12 named corporate partners and over 40 organizations through Project Glasswing, a defensive security program.
This is not the first access breach. A Discord group reportedly had unauthorized access to Mythos for two weeks before Anthropic discovered and cut it off.
Fable 5 is a consumer-facing version of Mythos with added safety guardrails. White House AI adviser David Sacks posted on X that Anthropic CEO Dario Amodei was told about the jailbreak but downplayed it.
Anthropic disputed this, saying the jailbreak was narrow and involved issues already present in other public models.
Sacks said the situation was “serious but easily resolved” and that the “ball is in Anthropic’s court.”
The suspension directly affects crypto and digital asset security. Coinbase had reportedly entered talks to access Mythos through Project Glasswing for defensive purposes.
Before the ban, Zcash’s founder claimed Mythos had audited the entire Zcash chain and found no bugs.
Security teams that relied on Mythos for vulnerability scanning now have no access. Anthropic has not announced when access will be restored.
The post China May Have Stolen America’s Most Dangerous AI — Here’s What Mythos Can Do appeared first on CoinCentral.
