OpenAI publishes a federal blueprint for frontier AI governance

OpenAI on June 3, 2026 published "A blueprint for democratic governance of frontier AI," a federal-level policy proposal that lays out how the U.S. government could build a durable institutional framework for governing increasingly capable AI systems. The blueprint accompanies a separate, broader public policy agenda published the same day, and reads as the more operational of the two documents.

What's new

The blueprint outlines a three-part strategy:

• Build a national framework that leverages emerging state consensus. OpenAI explicitly cites California SB 53, the New York RAISE Act, and Illinois SB 315 as the state-level templates the federal framework should build on. Each of those laws emphasizes transparency, public reporting around catastrophic-risk evaluations and safety incidents, whistleblower protections, and enforceable accountability for developers of the most capable models.
• Strengthen CAISI as the U.S. federal government's primary institution for frontier AI safety. OpenAI is proposing that the existing Center for AI Standards and Innovation become the durable institutional home for federal frontier-safety work, rather than spinning up a new body.
• Mobilize a broader resilience plan across government to address the national security and public safety challenges that flow from frontier AI, particularly cyber, chemical, biological, radiological, and nuclear (CBRN) risk.

OpenAI links the blueprint as a downloadable PDF on cdn.openai.com and cites a recent White House executive order on "Promoting Advanced Artificial Intelligence Innovation and Security" as the federal momentum the framework is meant to build on.

Context

The blueprint is the federal complement to a state-level pattern OpenAI has been endorsing for months. Both this document and the same-day public policy agenda name SB 53, the RAISE Act, and SB 315 as the acceptable state templates. The blueprint argues those laws have already done the hard work of building a transparency-and-reporting framework; the federal job, in OpenAI's view, is to harmonize and durably institutionalize what the states have proven out.

Routing the federal work through CAISI rather than proposing a new agency is a notable choice. It is the more politically tractable path — CAISI already exists, has staff, and has a mandate — and it tracks how OpenAI has previously argued frontier-safety work should be done.

Why it matters

This is OpenAI's most concrete federal-framework proposal to date and the first time the company has laid out, in one document, both the institutional home (CAISI) and the legislative template (the trio of state laws) it wants Washington to adopt. The proposal is consequential for two reasons.

First, by anchoring on CAISI, OpenAI is steering the federal conversation away from new-agency proposals and toward institutional reuse — a path that is faster but also concentrates safety oversight in a body that has not yet been tested against frontier-scale incidents. Second, by reaffirming the same three state laws across both documents, OpenAI is investing political capital in those frameworks specifically, which raises the cost for competing AI labs (or states) to pull federal policy toward different templates.

The blueprint itself is non-binding — it is a position paper, not legislation — but in the current Washington cycle, position papers from frontier labs often serve as the first drafts of staff-level bill text.