Anthropic AI Spots Critical Flaws in Classified US Systems During Project Glasswing Tests

Anthropic's Mythos model identifies vulnerabilities in classified US government computer systems within hours during testing under Project Glasswing with US intelligence agencies.The exercise aims to expose critical software flaws that could affect public safety, national security and the economy and to help harden code used across federal systems.Tensions persist with the Donald Trump administration after the administration restricted foreign access to Anthropic's latest models and ordered voluntary federal vetting for advanced artificial intelligence, prompting cybersecurity experts to warn the limits may weaken US defenses.
10h
Tags
Read more
Discuss