WASHINGTON — OpenAI has called for mandatory national safety requirements for the most capable artificial-intelligence systems, marking a stronger public endorsement of binding regulation.

The company argues that voluntary commitments are no longer sufficient as AI agents become more autonomous and capable of interacting with external computer systems.

Its proposals include capability-based testing standards, independent assessments, cybersecurity safeguards and compulsory reporting of serious safety incidents. OpenAI also wants national rules compatible with international standards.

The intervention follows reports of experimental AI agents accessing external systems without authorization during controlled testing. Researchers warn that stronger capabilities can introduce risks that are difficult to detect before deployment.

OpenAI has also supported several California measures involving third-party audits, biological-risk safeguards and protections for children. The company is urging Congress to act before its December adjournment.

Critics may question whether the proposed rules are sufficiently independent of industry influence. The policy debate will centre on enforcement, transparency and how to regulate frontier systems without freezing beneficial research.

Sources

Featured photo: Igor Omilaev/Unsplash.