AI
Sam Altman pushes back on Silicon Valley's vision of AI taking control
OpenAI's CEO warns that humans must remain the ultimate decision-makers as artificial intelligence grows more powerful, dismissing theories that machines should gradually take over key institutions.
TechCrunch — AI
OpenAI's leadership is drawing a clear line in the sand: as artificial intelligence becomes more capable, humans must remain firmly in control of critical decisions across society. Speaking at the New York Times Dealbook Summit this week, CEO Sam Altman and President Greg Brockman rejected the increasingly popular theory circulating through Silicon Valley that AI systems should gradually assume decision-making authority in major industries and institutions.
The handover debate#
Some technologists argue that as AI models become progressively more sophisticated and develop the ability to improve themselves—a process called "recursive self-improvement"—they will eventually evolve into a superintelligence beyond human control. Under this logic, a gradual transfer of power to these systems would be inevitable or even desirable.
Altman explicitly dismissed this framing. "There's been this thing in Silicon Valley that seems almost ridiculous to have to fight against, but we both believe very deeply that people need to stay in control of the future," he stated, referring to the theoretical handover concept as fundamentally misguided.
A shared conviction with technical disagreements#
While Altman and Brockman align on the core principle—that humans must retain ultimate authority—they differ on the technical approaches needed to achieve this goal. Rather than a fundamental philosophical divide, Altman characterized their debate as a conversation between two people working toward the same objective: preventing accidental or deliberate shifts in control.
"The question of how to actually prevent handover from happening...I think we can both agree on that," Altman explained, stressing the complexity of maintaining human oversight amid competitive and commercial pressures across the industry.
New monitoring challenges ahead#
As AI systems grow more sophisticated, Brockman highlighted an emerging technical hurdle: current monitoring methods may become inadequate. Today's researchers track AI reasoning by observing the model's "chain of thought"—the step-by-step explanations of how an AI solves problems. Yet more advanced models can accomplish complex tasks without explicitly writing out their reasoning, making this traditional oversight approach less effective.
"As the models have gotten more capable, they can do more within just the forward path without having to actually write out that thinking, which means we need a new approach to monitoring," Brockman said, calling these purely technical problems that engineers can solve.

