OpenAI CEO warns of AI alignment loss and power concentration as key risks

OpenAI chief executive Sam Altman said two scenarios could threaten the development of artificial intelligence: a loss of control over AI alignment and the concentration of excessive power in the hands of a single country or laboratory. In a post on X, he warned that if AI were to become misaligned, it would no longer serve humanity, and if an extraordinarily powerful system were controlled by one entity, it could impose a dystopian worldview.

Altman added that avoiding these threats requires a balanced approach, noting that both governmental and corporate dominance pose risks. He reiterated OpenAI’s support for a federal framework that would set consistent safety standards for frontier AI and said the company is adopting new safety processes, including explicit safety cases before major reinforcement-learning experiments.

"We are unapologetically on Team Humanity, and AI must always serve people," Altman wrote.

"Committing to having independent evaluators with employee-like access is a great idea, and we will do the same. We'll have more to share soon," he added, echoing calls from other AI leaders for third-party oversight.

Altman’s remarks come amid broader industry discussions about ensuring AI alignment as models become more capable, with researchers warning of existential risks if safety measures lag behind technical progress.