Anthropic’s Mythos AI Reportedly Hacked the NSA’s Most Sensitive Systems ‘in Hours’
Anthropic's Mythos AI model reportedly accessed highly sensitive U.S. government and NSA systems within hours during a security exercise. This event preceded a U.S. government ban on foreign access to the company's top-tier models. The crackdown has led to lawsuits from a legal tech firm and an Anthropic customer.
What changed
Reports now link a specific red-team evaluation on June 11 to the subsequent U.S. export ban.
Live updates
-
Anthropic Mythos AI Reportedly Breached NSA Systems During Testing
confidence 80%Anthropic's Mythos AI model reportedly accessed highly sensitive U.S. government and NSA systems within hours during a security exercise. This event preceded a U.S. government ban on foreign access to the company's top-tier models. The crackdown has led to lawsuits from a legal tech firm and an Anthropic customer.
What's confirmed:
- A U.S. official stated an Anthropic AI model identified vulnerabilities in highly sensitive U.S. government computer systems during testing.
- The U.S. government has banned foreign nationals from accessing Fable 5 and Mythos 5.
- The U.S. export ban on Anthropic AI models has affected South Korea.
- A legal tech firm and an Anthropic customer have filed lawsuits against the U.S. over the loss of access to AI models.
Still unconfirmed:
- Mythos AI infiltrated nearly all NSA classified systems within a few hours during an authorized red-team evaluation on June 11.
- A phone call from Amazon CEO Andy Jassy triggered the chaos of the Anthropic crackdown.