US Breached: Advanced Weapons Supplier Exposes Government Weaknesses

AI company Anthropic has attempted to identify sensitive vulnerabilities within the United States government. Testing with its Mythos model revealed many hidden secrets.

According to the Associated Press, the testing was conducted under Project Glasswing, a limited programme designed to find and fix critical software vulnerabilities before they could be attacked.

Mythos successfully uncovered numerous weaknesses in US government systems, accomplishing the breach in a very short time.

“Mythos broke into almost all classified systems in a matter of hours, not weeks,” Senator Mark Warner said, as quoted by Reuters on Wednesday.

However, according to an official, this identification does not mean Mythos could exploit US security simultaneously.

Neither the White House, the Department of Defense, nor Anthropic immediately responded to requests for comment on the report.

Anthropic’s relationship with the US government has not always been smooth. The two parties are known to have collaborated so that the company’s Claude data could be used by the government in daily operations.

However, that relationship subsequently soured after Anthropic refused to allow the US military to use its AI models for surveillance and fully autonomous weapons systems.

Washington did not remain silent and retaliated. The Trump administration placed Anthropic on a national security blacklist.

Furthermore, the US government ordered the suspension of exports of Anthropic’s Mythos and Fable models to various countries worldwide, citing national security concerns.