OpenAI CEO Sam Altman has warned that the rapid development of artificial intelligence could take a dangerous turn, identifying two scenarios he believes the industry must prevent as increasingly powerful systems are built.
Altman outlined his concerns in a post on X late Sunday, shortly after former AI researcher Jacob Coxon publicly accused major AI companies of moving too quickly toward advanced systems without adequate safeguards.
Altman Warns Against Losing Control of AI
According to Altman, the first major danger would be humanity losing control over the future direction of artificial intelligence.
“We could lose control of the future of AI,” he wrote, describing that possibility as “unacceptable.”
The OpenAI chief stressed that his company remains firmly focused on ensuring AI serves humanity rather than operating beyond meaningful human oversight.
“We are unapologetically on Team Humanity and AI must always serve people,” Altman said.
He argued that safety and alignment research must advance quickly enough to keep pace with improvements in AI capabilities.
For Altman, the issue is not whether AI development should continue, but whether safeguards can remain ahead of increasingly capable models.
Concentrated AI Power Raises a Second Concern
Altman identified a second potential danger: excessive control of powerful AI being concentrated in the hands of a single individual or company.
He warned that an extraordinarily powerful AI system controlled by one person or organisation could potentially be used to impose that entity’s worldview on society.
Such a situation, he suggested, could create a deeply dystopian future in which a small concentration of power determines how advanced AI is used.
The warning reflects a broader debate within the technology industry over who should control increasingly sophisticated AI systems and how much authority should be given to the companies developing them.
OpenAI Chief Calls for a Slower Pace
Altman has also sought to clarify what he means when he calls for AI development to be “paced.”
In an earlier X post, he stressed that slowing progress does not mean abandoning technological advancement.
“When we talk about ‘pacing,’ we don’t mean ‘stopping,’” Altman wrote.
“Progress has been rapid and will continue to be. But it should be slower than it otherwise could be,” he added.
He argued that a more measured approach would give researchers additional time to address safety concerns before AI capabilities move significantly further ahead.
Altman Rejects Competition as an Excuse for Recklessness
The OpenAI CEO acknowledged the competitive pressure surrounding the development of advanced AI, particularly among American technology companies.
However, he argued that competition should not become a justification for cutting corners on safety.
“Pacing will be well worth the cost,” Altman said, adding that American competitive pressure should not allow AI capabilities to move ahead of alignment and monitoring.
He also called on every leading AI laboratory to demonstrate that increasingly powerful systems can be developed responsibly.
“The world deserves confidence that American companies developing increasingly capable AI will act responsibly,” he wrote.
Whistleblower Jacob Coxon Raises Alarm
Altman’s comments came shortly after Jacob Coxon announced his resignation from Anthropic, where he had worked on pretraining research.
Coxon, who said he had previously worked at both Anthropic and OpenAI, used social media to accuse the companies of failing to approach the development of advanced AI responsibly.
“I resigned from Anthropic today,” Coxon wrote on X.
“Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives.”
His warning centred on the possibility of AI systems eventually becoming capable of improving themselves and operating at levels beyond human intelligence.
Coxon Warns Against Underestimating Advanced AI
Coxon urged the public not to underestimate the potential capabilities of future AI systems.
He argued that increasingly advanced systems could eventually become capable of activities such as sophisticated cyberattacks, rapidly transforming scientific and other fields, and accumulating significant real-world resources and influence.
He also claimed that some people working directly on AI privately hold much more serious concerns than they express publicly.
According to Coxon, senior executives and researchers sometimes use more restrained language when speaking publicly, despite privately discussing the possibility of severe risks from advanced AI.
Geoffrey Hinton Has Issued Similar Warnings
Coxon’s concerns echo warnings previously made by Geoffrey Hinton, the Canadian computer scientist widely known as the “Godfather of AI.”
Hinton has argued that developing superintelligent AI without a clear scientific understanding of how to keep such systems safe and controllable could be extremely dangerous.
“We would be very foolish to develop superintelligence now, when there is no scientific consensus it can be developed safely and controllably,” Hinton has said.
He has also warned that losing control over an AI system more intelligent than humans could have catastrophic consequences, including the possibility of human extinction.
