AI threats and model safety dominate the day: researchers demo a self-replicating local-model worm, while vendors race to improve security guardrails for new AI models. Defenders face increasing pressure to manage agent behavior, privacy risks, and deepfake threats on social platforms.