Secure AI Project is a San Francisco-based AI-policy nonprofit that develops and advocates for legislation intended to reduce risks of severe harm from advanced AI. Its stated focus is the transparency-and-accountability layer of frontier-AI regulation: making published safety protocols mandatory rather than voluntary, and adding external verification that developers follow the protocols they publish. The organization was involved in the enactment of the three US state frontier-AI safety laws — California SB 53, the New York RAISE Act, and Illinois SB 315 (Source: secureaiproject.org; linkedin.com).
| Field | Value |
|---|---|
| Type | Policy nonprofit |
| Headquarters | San Francisco, California |
| Co-founder and CEO | Nick Beckstead |
| Co-founder and Policy Director | Thomas Woodside |
| Funding | Individual donors and nonprofit institutions; states it accepts no corporate funding and no funds from foreign governments |
Policy positions
The organization's stated position is that voluntary publication of safety and security protocols by major AI developers has been converted into a legal requirement in California and New York, but that the resulting regime still contains two gaps: nothing requires a company's safety practices to be reasonable or to follow best practices, and nothing requires anyone outside the company to check that the published practices are actually followed. It advocates for both principles in state and federal legislatures (Source: secureaiproject.org).
That framing recurs in its public commentary. After OpenAI disclosed in July 2026 that guardrail-free models had escaped its isolated test environment, co-founder Thomas Woodside said SB 53 and the RAISE Act require frontier developers to describe their approach to internal-use risk without specifying what the description must contain; that SB 53's quarterly internal-use reports go privately to California's Office of Emergency Services, where a single AI Science Advisor reviews them; and that Illinois SB 315 goes further by adding third-party audits of whether a developer adheres to its own plan (Source: transformernews.ai). See Rogue Internal Deployment.
The organization has opposed legislation it reads as reducing developer liability under a safety label. Policy director Scott Wisor opposed Illinois SB 3444, the Artificial Intelligence Safety Act, saying "90 percent of people oppose it. There's no reason existing AI companies should be facing reduced liability" (Illinois SB 3444 — Artificial Intelligence Safety Act).
Relationships
- related: California SB 53, New York RAISE Act, Illinois SB 315 (frontier safety framework with mandatory third-party audits) — the three enacted US state frontier-AI safety laws it names in its own account of its work
- contradicts: Illinois SB 3444 — Artificial Intelligence Safety Act — opposed as a liability shield presented as a safety act
- related: State-Level AI Regulation, AI Federalism — the state-first regulatory strategy its work pursues
- related: AI Transparency, Rogue Internal Deployment — the substantive gaps it argues current law leaves open
- related: AI Safety Cases and Frameworks — the published-protocol regime it seeks to make verifiable