OpenAI CEO Sam Altman Admits AI Agents Are Becoming A Serious Risk

OpenAI CEO Sam Altman has openly admitted that AI agents are starting to become a serious problem, especially as they grow powerful enough to create real-world risks. In a recent post on X, Altman warned

that advanced AI agents can now find critical security flaws which could be misused if left unchecked."We have a strong foundation of measuring growing capabilities but we are entering a world where we need more nuanced understanding and measurement of how those capabilities could be abused, and how we can limit those downsides both in our products and in the world, in a way that lets us all enjoy the tremendous benefits. These questions are hard and there is little precedent; a lot of ideas that sound good have some real edge cases," he wrote on Elon Musk-owned social media platform.

Altman said that AI models have improved very quickly in the past year and can already do many impressive things. But this rapid progress also brings new problems. According to him, AI agents are now reaching a level where they can not only help experts but also be misused by attackers if proper safeguards are not in place.OpenAI Is Hiring For Head Of Preparedness Role To Stay A Step Ahead Of AI In Maintaining Safety Standards

He explained that OpenAI is hiring a Head of Preparedness to handle this challenge, as the company needs a deeper understanding of how advanced AI capabilities can be abused. "We are seeing models become good enough at computer security that they are beginning to find critical vulnerabilities," Altman wrote.His comments come at a time when the global threat from AI-powered cyberattacks is rising. Recently, Anthropic reported that Chinese state-sponsored hackers misused its Claude Code tool to target around 30 organisations, including tech firms, banks and government bodies. The attacks reportedly required very little human involvement, raising concerns about how easily AI agents can automate hacking attempts.Altman also mentioned that AI’s impact is not limited to cybersecurity. He said OpenAI had already seen early signs of mental-health risks from AI interactions back in 2025. Several lawsuits and reports have accused AI chatbots of spreading misinformation.