OpenAI has released a blueprint calling for a new U.S. federal body to regulate frontier AI systems, according to a post on the company's website. The proposal seeks to establish a regulatory framework focused specifically on the most advanced models, arguing that existing laws are not equipped to address risks posed by systems that exceed current capabilities.
The Three Pillars of Governance
OpenAI's blueprint rests on three areas:
- Safety and Security Standards: Mandatory pre-deployment risk assessments, independent audits, and security protocols aimed at preventing model theft or misuse. The proposal calls for mandatory, independent audits for any AI model exceeding a specific capability threshold.
- National Security: Controlling access to the most powerful models to prevent acquisition by malicious state actors or terrorist groups.
- Societal Resilience: Investment in defenses against AI-enabled threats such as disinformation and cyberattacks, including tools for content authentication and public education efforts.
Why It Matters
OpenAI's proposal is notable because it comes from one of the industry's leading developers rather than from regulators or outside advocacy groups. It gives lawmakers a concrete starting point as Washington considers how to legislate frontier AI, though it remains a proposal from an interested party rather than an enacted framework.