One of Anthropic's most capable artificial intelligence models, Mythos, identified vulnerabilities in highly sensitive and secure classified US government computer systems. A US official stated that the flaws were uncovered during a testing exercise.
The model surfaced the vulnerabilities across the systems in mere hours rather than weeks. The official noted that finding a weakness within hours is not the same as exploitation.
Earlier this month, Trump directed the company to stop foreign nationals from using its latest AI models.