they even published papers where they distill gpt models… https://alignment.anthropic.com/2025/subliminal-learning/ https://alignment.anthropic.com/2025/sublim

they even published papers where they distill gpt models… https://alignment.anthropic.com/2025/subliminal-learning/ https://alignment.anthropic.com/2025/subliminal-learning/ anthropic is really adversarial.