OpenAI released a blog post on September 21 stating that the United States should lead an international effort to establish technical standards for frontier AI development. The company emphasized the need for global cooperation to address challenges such as regulatory fragmentation, the need for collective action, and uneven global capacity.
OpenAI Calls for US-Led International Standards for Frontier AI and Recursive Self-Improvement
A key focus of the proposal is managing the risks associated with recursive self-improvement (RSI)—a process where AI research becomes increasingly automated, potentially accelerating the pace of AI progress. While OpenAI noted that fully autonomous RSI is not currently occurring, the company warned that without robust safeguards and human oversight, such processes could lead to a loss of practical control over AI development.
The company suggested leveraging existing networks of AI safety institutes, including those in Japan, the UK, Canada, France, Germany, South Korea, Singapore, India, Australia, and Kenya. Proposed standards would cover capability benchmarks, risk management for automated research, and incident classification. OpenAI clarified that these standards should serve as a technical foundation for safety rather than as mandatory licenses or approval requirements, leaving legal implementation to individual national governments.
Sources
- OpenAI、フロンティアAIの国際標準づくりを米国主導で進めるよう提言 「RSIはまだ起きていない」 （ITmedia AI+, 2026-09-22）
- 公式ブログ