OpenAI wants US to lead global standards for AI self-improvement

OpenAI wants the US to lead an international effort to set technical standards for frontier AI, including recursive self-improvement (RSI), as AI systems take on more of the work involved in developing the next generation of AI.

In a blog post outlining its priorities for the next phase of AI development, OpenAI said automated AI research could significantly accelerate the pace of AI progress. But it warned that fully autonomous RSI should not be pursued unless humans can retain meaningful control over the process.

“Fully autonomous RSI is not happening today, and we should not pursue it unless and until it can be done safely,” the company said.


RSI refers to AI systems taking on increasing amounts of the work involved in developing successive generations of AI. As more of this process becomes automated, OpenAI said, the pace of progress could accelerate rapidly.

The company sees benefits in this, including lowering the cost of advanced intelligence, speeding up scientific discoveries and helping with AI safety research. An automated AI researcher could also become an automated AI safety researcher, OpenAI said.

Also Read: DeepSeek to brief UN Security Council on AI this week, sources say

But the risks could grow just as quickly.

If AI systems begin driving their own development without adequate safeguards, humans could lose practical control over the process and struggle to understand research being carried out by systems that are increasingly autonomous.

“From here, AI could become more dangerous, less aligned, and, on the whole, a danger to people,” OpenAI said.

OpenAI said deciding whether and how to move towards greater automation must depend on whether human control can be preserved, as well as on informed democratic choices about the benefits and risks.

OpenAI wants common AI safety standards

OpenAI said alignment research needs to keep pace with AI capabilities, but argued that technical research alone will not be enough. It wants countries to develop common standards for frontier AI safety and security.

These standards could establish shared ways of measuring AI capabilities, evaluating risks and determining whether safeguards are sufficient. They could also make it easier for countries to compare evidence and respond to AI incidents that cross borders.

OpenAI identified fragmented national evaluation and reporting systems, collective-action problems and uneven distribution of frontier AI expertise as key challenges.

Its proposal is for the US to work with countries around the world to develop global technical standards for frontier AI, including RSI.

OpenAI suggested using the growing network of national AI safety institutes, including those in India, the UK, Japan, South Korea, Singapore, France, Germany, Canada, Australia and Kenya, to help develop these standards.

The standards could cover frontier AI capability benchmarks, risk assessment, human oversight of automated AI research and incident reporting.

Also Read: AI leaders called for a slowdown; now they're being sued over it

OpenAI stressed that these would not be licences or mandatory pre-release approval systems. Individual governments would decide whether and how to incorporate them into their own laws.

The company also said the process should include open-model and closed-model developers, independent researchers and academia, and should not favour particular companies or countries.

OpenAI also called for secure channels between governments and critical infrastructure operators to share information on national security concerns, vulnerabilities and emerging threats related to frontier AI.

It said greater dialogue between the US and China on AI safety would be a positive step.

For OpenAI, “pacing” AI development does not mean setting a fixed speed for technological progress. Instead, it means ensuring alignment research and safety measures stay ahead of AI capabilities.

The company said the US is well placed to lead this effort because its AI industry remains at the technical frontier and because of its influence across finance, trade, defence, technology and information systems.

“Leading now will determine whether the United States shapes the global AI framework or watches a fragmented, uneven, and conflict-ridden system take hold around it,” OpenAI said.