US Builds AI Safety Framework

The United States is forging a national standard for AI safety through a dual approach involving state-level action and federal initiatives, aiming to establish a US-led global framework. This strategy, termed 'reverse federalism' by OpenAI, sees states aligning on AI safeguards, creating a foundation for federal policy.

California, New York, and Illinois have passed frontier AI safety legislation, establishing common baselines for governing powerful AI systems. These state efforts are converging with federal work, shaping a national standard.

OpenAI argues that critical decisions about AI, especially frontier safety, should be democratically governed. A national framework, grounded in this principle, is ultimately ideal.

In the interim, states are creating a de facto national standard by passing mirroring laws. This approach aims to foster innovation while preventing a fragmented regulatory landscape that could hinder development and global competitiveness.

The core elements identified across these state laws include documented safety frameworks with risk assessments and public disclosures, reporting of serious safety incidents, and governance through independent audits. These measures aim to build democratic oversight into AI deployment.

However, states should not be tasked with managing national security risks or conducting highly technical reviews better suited for federal experts. Such responsibilities require federal resources, expertise, and access to classified systems.

Federal Action

At the federal level, the Trump Administration is developing a framework for testing advanced AI models, focusing on cybersecurity. OpenAI is actively participating in discussions with the administration and other stakeholders to shape this effort.

Consistency in testing standards and processes is paramount. The current, often ad-hoc, testing of AI models highlights the need for a repeatable approach at both state and federal levels.

A federal testing framework is expected to facilitate the deployment of advanced AI tools to government, critical infrastructure, and allied partners, thereby strengthening democratic institutions.

Congress is also advancing proposals for a federal framework, with bipartisan support emerging for national governance and frontier safety measures. Incorporating successful state approaches into federal legislation could streamline the establishment of a unified national regime.

OpenAI's blueprint for frontier safety includes federal leadership in testing and evaluation, requiring clear standards for companies such as independent audits and incident reporting, and ensuring federal and state efforts reinforce each other.

Global Implications

National legislation is also critical for establishing a US-led international framework for AI standards. Discussions at the G7 and subsequent proposals from AI leaders underscore the growing need for global consensus.

A bipartisan federal framework would provide a robust foundation for international collaboration on AI standards, risk assessment, and responsible technology deployment.

The convergence of state alignment, federal development, and international dialogue positions the US to lead in shaping a global AI ecosystem grounded in democratic values.