OpenAI calls for US to lead effort to develop technical standards for advanced AI

OpenAI has called on the United States to take the lead in developing global technical standards for advanced artificial intelligence. CEO Sam Altman argued that increasingly capable AI systems require common approaches to safety, evaluation and risk management across countries.

The proposal comes as governments and technology companies face growing questions over how to manage so-called frontier AI systems, particularly models that can operate with greater autonomy and improve their own capabilities. OpenAI’s call was made ahead of CEO Sam Altman’s expected engagement with the United Nations on AI-related issues.

In a policy post, OpenAI said that national rules alone may not be sufficient because AI models, research and technical expertise operate across borders. The company is calling for compatible international approaches covering areas such as measuring AI capabilities, managing risks, maintaining human control and determining when development should slow or stop.

One area receiving particular attention is recursive self-improvement. This is a scenario in which AI systems could autonomously enhance their own capabilities. OpenAI has identified the possibility as an important consideration for future AI governance and safety efforts.

What standards could cover

OpenAI’s proposal focuses not only on government regulation but also on technical practices within companies developing frontier AI systems. The company has called for stronger monitoring of AI models, particularly for potentially misaligned behaviour, where a system acts in ways that conflict with human intentions or established boundaries.

The company has also advocated for reporting mechanisms for serious AI incidents and greater transparency around how advanced systems are evaluated. OpenAI says industry-led standards could provide an immediate framework.

The proposal follows a series of recent efforts by OpenAI to establish frameworks for AI safety. Earlier this month, the company introduced a framework for tracking and publicly reporting model-misalignment incidents.

OpenAI has also previously advocated for national AI safety standards in the US and described state-level efforts as potentially laying the groundwork for a broader national framework.