Mythos Model Unveils U.S. Government Cyber Vulnerabilities

Mythos, Anthropic's AI model, revealed vulnerabilities in U.S. government computer systems during a classified test under Project Glasswing. Despite its success, Anthropic's relations with the U.S. government have been strained due to restrictions on military use of its AI technology, leading to a national security blacklist and export bans.

In a remarkable turn of events, Anthropic's AI model, known as Mythos, exposed critical vulnerabilities in U.S. government computer systems within hours during a testing exercise under the auspices of Project Glasswing, multiple sources reported. This exercise, undertaken in collaboration with Washington's intelligence agencies, aims to detect and rectify software loopholes before they can be exploited.

The Associated Press, citing Senator Mark Warner, noted his remarks in a congressional hearing where he relayed information from NSA chief Joshua Rudd about Mythos's testing capabilities. Senator Warner emphasized that Mythos had breached nearly all classified systems, illustrating the model's capabilities to identify, though not necessarily exploit, vulnerabilities in rapid time.

However, Anthropic's path with the U.S. government has seen friction. The company's refusal to allow its AI for domestic surveillance and autonomous military systems led to its national security blacklist status. Furthermore, export of Anthropic's Mythos and Fable models faces suspension due to ongoing national security concerns, as highlighted by recent government actions and reports of restricted NSA access to Mythos.