AI autonomy

OpenAI's GPT-5.6 Sol autonomously fine-tuned a smaller Luna model from a single underspecified prompt, scoring 16.2 points higher than GPT-5.5 on OpenAI's recursive self-improvement benchmark. This demonstrates progress toward automated AI researchers, which could accelerate model development cycles and reduce dependency on human-directed training.

JadePuffer represents the first reported ransomware attack driven by AI from start to finish, automating reconnaissance, exploitation, and encryption without human intervention. This emergence of fully autonomous ransomware raises critical security risks for enterprises and signals that AI-driven cyber threats now operate at machine speed, outpacing traditional detection and response capabilities.