OpenAI's Sam Altman backs frontier AI safety rules, says 'pacing' development does not mean stopping

OpenAI CEO Sam Altman has called for shared safety standards across frontier AI companies, saying the industry must build confidence that increasingly capable systems are being developed responsibly.

In a post on X, Altman welcomed a federal framework with consistent safety requirements for frontier AI, including the possibility of independent auditors. But he said companies should not wait for legislation or an antitrust exemption to begin the work.

“Every frontier lab must deliver on this,” Altman said, referring to the need for responsible development. “There is no reason any of us should come to work if we cannot.”


His comments come days after leading AI executives, including Altman and Anthropic CEO Dario Amodei, backed the idea of pacing frontier AI development amid concerns that increasingly capable systems could become difficult to control.

Altman said there were “two ways AI progress could go very badly” that the industry must avoid.

“First, we could lose control of the future to AI. This is unacceptable; we are unapologetically on Team Humanity, and AI must always serve people,” he said. Ensuring that, he added, would require alignment and safety techniques to remain ahead of progress in model capabilities.

The second risk, according to Altman, is excessive concentration of power.

“We could end up in a world with too much concentration of power. If an extraordinarily powerful AI is used by one person or company to impress their worldview onto everyone else, the results could be extremely dystopian,” he said.

Avoiding these risks would require “walking a narrow middle path”, Altman said. That could include preventing one country from gaining too much power, as well as avoiding a situation where one AI lab becomes disproportionately powerful.

Also Read: Is it possible to slow down the development of AI?

From model releases to development risks

Altman said OpenAI’s earlier Responsible Scaling Policies and Preparedness Frameworks had focused mainly on completed models and their deployment. The company is now also looking at risks during the development process.

OpenAI is formulating explicit safety cases before frontier reinforcement learning runs that are expected to significantly increase a model’s capabilities, in addition to its pre-release safety work.

Also Read: Anthropic CEO Dario Amodei calls for slower AI progress, warns of 'catastrophic' risks

Altman said he hoped other AI companies would share their approaches so that the industry could develop common standards around misalignment, monitoring and safety.

He also clarified that “pacing” AI development does not mean stopping progress altogether. Development would continue, he said, but could be slower than it otherwise might be because of the additional work required for safety cases and monitoring.

“No amount of US competitive pressure justifies recklessness or capabilities getting ahead of alignment and monitoring,” Altman said, adding that governments would be needed for international coordination but that the industry should begin taking steps on its own.