OpenAI Proposes Global Safety Standards for Advanced AI

OpenAI has proposed a framework for global standards on advanced AI safety. The proposal calls for binding rules to govern recursive self-improvement, human oversight, risk assessment, and incident reporting.

The framework would cover areas where national rules alone may not be enough to address risks crossing borders. OpenAI argues that existing national institutions could coordinate technical standards across jurisdictions.

The company points to the Center for AI Standards and Innovation as a possible anchor for U.S.-led international standards. It also cites safety institutes operating in Europe, Asia, Africa, and North America.

OpenAI's proposal does not create licenses or mandatory prerelease approvals. Instead, national governments would decide how technical standards are incorporated into domestic law.