Science ❯ Computer Science ❯ Machine Learning ❯ Reinforcement Learning
The model can autonomously chain and exploit software flaws, forcing OpenAI to limit powerful cyber features to vetted defenders and add new controls.
