The US AI industry is looking to slow the development of increasingly advanced AI systems amid growing concerns that more powerful models could eventually threaten humanity. OpenAI CEO Sam Altman has argued that companies cannot afford to act recklessly and should take responsibility themselves rather than waiting for the Trump administration to introduce new policies.
“We welcome a federal framework that sets consistent safety requirements for frontier AI,” Altman wrote on X. “But we do not believe we need to wait for an anti-trust exemption or legislation to begin the work of providing this confidence.”
Altman also said that “no amount of American competitive pressure should justify recklessness, or let capabilities get ahead of alignment and monitoring.” While he acknowledged that government involvement would be important for international coordination, he stressed that companies should first take whatever steps they can on their own.
The OpenAI chief's comments came shortly after US President Donald Trump dismissed the need for regulating frontier AI development. Trump recently said that while certain safeguards could be introduced, he believed some of the concerns being raised were unlikely to materialise. He also said he wanted to maintain the current approach because “whoever wins AI wins.”
In his post, Altman said every frontier AI laboratory had a responsibility to behave safely and argued there was “no reason any of us should come to work if we cannot.” He also expressed support for independent auditors, an idea previously proposed by Anthropic CEO Dario Amodei.
New standards needed for AI safety, says Sam Altman
Although AI companies have previously introduced safety approaches such as Responsible Scaling Policies and Preparedness Frameworks, Altman said the industry now needs new mechanisms focused specifically on safe development and evaluation.
“At OpenAI we now formulate explicit safety cases in advance of frontier reinforcement learning runs we expect to significantly increase capability, in addition to the safety work we have long done in advance of model releases,” he said.
Altman said OpenAI hoped other AI companies would study its approach and develop their own systems. He argued that common standards covering misalignment, monitoring and safety could produce better outcomes and said the company was willing to work with others across the industry to develop the strongest possible framework.
Addressing the idea of “pacing” AI development, Altman clarified that it did not mean bringing progress to a halt. AI development has been moving rapidly and will continue to do so, he said, but it should proceed more slowly than it otherwise might because measures such as safety evaluations and monitoring involve substantial costs.
Sam Altman outlines two nightmare AI scenarios
In a response to his original post, Altman described two potential nightmare scenarios involving AI. The first involves humanity completely losing control over artificial intelligence, something he said the industry must prevent.
“This is unacceptable; we are unapologetically on Team Humanity, and AI must always serve people,” he wrote. He added that ensuring this would require methods capable of keeping alignment and safety techniques ahead of advances in AI capabilities.
The second scenario involves an individual or company gaining excessive control over AI and using that power to impose its worldview on society, potentially creating a dystopian future.
Altman said avoiding both outcomes would require maintaining a delicate balance so that neither a single country nor a single company gains disproportionate control over AI.
“Avoiding these two threats requires walking a narrow middle path,” the OpenAI CEO wrote, pointing to the possibility of one country or one AI laboratory becoming too powerful.
AI safety has become a major focus in the technology industry after Anthropic CEO Dario Amodei urged companies to slow the pace at which AI capabilities are advancing. Amodei proposed a three-part framework involving independent evaluators with employee-like access, cooperation between leading AI companies on safety standards and greater international coordination. Altman supported the proposal, saying that giving independent evaluators employee-like access was a good idea and that OpenAI would adopt the same approach. Elon Musk also backed Amodei's position, replying, “Dario is right.”
At the same time, while AI companies have been calling for a federal regulatory framework in the US, former White House AI czar David Sacks argued that leading AI companies did not need government permission to slow development. “Stop pretending you need anyone else’s permission. Stop pretending antitrust law has to be suspended so you can form a cartel,” he wrote on X.
Meanwhile, researchers at major AI companies, including Anthropic and OpenAI, have voiced concerns about the potential consequences of increasingly powerful AI systems. Anthropic researcher Jacob Coxon resigned from the company, claiming that people developing AI genuinely believe the technology could potentially kill everyone by the end of the decade. OpenAI safety researcher Marcus Williams, meanwhile, estimated that there was a 70 per cent possibility that AI could cause human extinction within the next three years.
