Sam Altman Says Agentic AI Is the Most Consequential Safety Challenge Yet

Sam Altman said agentic AI represents the most consequential safety challenge yet, arguing the stakes change dramatically once AI can act directly inside real systems.
- People may take time to get comfortable with agentic AI, just as they once did with putting credit cards online
- AI systems will increasingly be able to click around the internet and interact directly with computers and services
- Once AI has access to systems, information and the ability to take actions, mistakes become much higher stakes