Zvi published a Substack post discussing further developments involving internal AI models that hack, or attempt to hack, things. The available record contains only the post title and Hacker News metadata: the item had a score of 2 and 0 comments. It does not provide model names, incident details, technical mechanisms, evidence, or independent verification, so the claims cannot presently be assessed in depth.
No heat snapshots are available in the last 24 hours.