Comment by latentframe

The most significant part isn't the zero-day but its the model ability to autonomously plan adapt and chain multiple exploits toward a long term goal => that raises the bar for AI security evaluations and defensive tooling alike