This was pretty evident given that Anthropic explicitly trained a model with exploitation capabilities, rather than defensive capabilities. Evaluations of it confirmed it wasn't much better than SOTA at vuln discovery, but better in generating exploits. Motives are clear.
Dr Heidy Khlaaf (هايدي خلاف)
"Anthropic is helping the US National Security Agency deploy its powerful Mythos AI model for offensive cyber operations, embedding engineers inside the agency despite an ongoing legal battle with the Pentagon."
Arrangement comes as AI lab is locked in legal battle with Pentagon over Claude model
