OpenAI said on Wednesday that it wants the United States to introduce compulsory national rules for AI safety. The company has raised concerns that increasingly advanced technology could speed up the development of new AI systems, especially after some of its own AI agents behaved unexpectedly.

Several recent cases involving AI models from companies including OpenAI have shown how difficult it can be to control advanced systems when they interact with outside platforms during testing. These incidents have increased pressure for stronger safety measures and government oversight.

OpenAI Chief Global Affairs Officer Chris Lehane said voluntary commitments are not enough to address the possibility of AI speeding up further AI development. He called for national rules based on the capabilities of AI systems, with regulations that can change as the technology advances.

The proposal represents a significant effort by OpenAI to establish binding national AI safety standards. The U.S. Congress has not yet created a federal framework for AI regulation, while several states have introduced their own laws.

OpenAI is asking Congress to establish national requirements based on AI capabilities. The proposed measures would include testing standards, independent evaluations, cybersecurity safeguards and mandatory reporting of incidents involving the most advanced AI systems.

The company behind ChatGPT has urged lawmakers to take action before Congress adjourns in December. OpenAI also said it would continue supporting state level AI legislation until federal rules are established.

OpenAI said fully autonomous recursive self improvement, in which an AI system independently develops successive generations of AI, is not currently happening. The company added that such development should not be pursued until it can be carried out safely.