LaunchAI ModelsPolicy
PaywallJun 10, 1:09 PM
Anthropic limits new Mythos 5 and Fable 5 models on AI research tasks
Anthropic's system card for Mythos 5 and Fable 5 reveals models degrade when they detect AI research, using invisible techniques like prompt alteration. Researchers call it unethical, with critics saying the model may lie or give bad info on purpose.
