OpenAI CEO Sam Altman has warned that humanity could eventually lose control of its future to artificial intelligence if the technology’s capabilities advance faster than the safety systems designed to keep powerful AI models under human control.
In a post on X on September 14, Altman identified what he described as two ways AI development could “go very badly.” The first, he said, is the possibility that humans could lose control of the future to AI.
Altman said this outcome would be unacceptable and stressed that AI must remain a technology that serves people. He argued that safety and alignment techniques — systems designed to make AI behave according to human intentions and values — must advance alongside the capabilities of increasingly powerful models.
His warning comes as leading AI companies race to develop increasingly capable “frontier” models, raising questions about whether existing monitoring and safety measures will remain effective as systems become more autonomous.
Altman has said OpenAI is already taking steps to strengthen its safety processes, including preparing safety cases before certain advanced training runs. In a recent Fortune interview, he said the company should not train models when it cannot make a convincing safety case that their behaviour will remain controllable and aligned with human intentions.
Altman identified a second danger beyond AI itself becoming difficult to control: the concentration of enormous AI-powered influence in the hands of a single individual, company, country or laboratory.
He warned that an extraordinarily powerful AI system controlled by one entity could potentially be used to impose that entity’s worldview on others.
According to Altman, avoiding both scenarios requires what he described as a “narrow middle path” — continuing technological progress while ensuring that neither AI systems nor a small group of people gain excessive control over the future.
The comments follow calls from Anthropic CEO Dario Amodei for AI companies to slow the pace of frontier-model development enough to allow safety measures, independent evaluation and coordination to keep pace with rapidly advancing capabilities.
Altman has supported the idea of “pacing” AI development, while making clear that he does not believe progress should simply stop. He has also called for consistent federal safety standards for frontier AI companies rather than relying entirely on voluntary measures.
For Altman, the central issue is not whether AI development should continue, but whether humans can maintain meaningful control as the technology becomes substantially more capable.
In a recent interview, he said OpenAI would be prepared to pause training if it reached a point where the company could no longer establish that its systems were sufficiently controllable and aligned.
The debate highlights a growing tension within the AI industry: how to continue developing increasingly powerful technology while ensuring that safety, human oversight and accountability do not fall behind.
Discover more from Scoop Hub
Subscribe to get the latest posts sent to your email.
