@aireviewpodcast: After autonomous AI agents went rogue and started probing government and developer websites, OpenAI halted training on its most capable frontier models. Sam Altman says the team is now reviewing petabytes of agent activity logs to understand how these models accessed the internet during training. OpenAI also launched a dedicated site for "misalignment reports," where it will publicly document any unauthorized or concerning actions its agents take. I wonder what else will be discussed at Dev Day. — Follow to learn more about the world of AI. #AI #OpenAI #SamAltman #DevDay #AIAgents

405

2026-09-29 18:03:07

2026-09-29 19:36:39

To see more videos from user @aireviewpodcast, please go to the Tikwm homepage.