An independent nonprofit that defines safety standards for frontier AI companies and assesses their adherence to them.
Guidelight AI Standards is an independent nonprofit that works to define and promote important safety practices for companies developing frontier AI models. It operates through three functions: (1) Standards โ pairing safety principles with concrete practices; (2) Assessments โ evaluating how well companies adhere to the standards and publishing the results based on public information; and (3) Incentives โ supporting internal safety advocates and building external stakeholder pressure to drive wider adoption of safer practices. Guidelight has published three standards โ Control, Capability Testing, and Transparency โ and an assessment report, 'AI Control: An Assessment of Frontier Practices' (18 August 2026). The organization was founded by Page Hedley and Steven Adler, both former OpenAI safety leaders. Guidelight states that it is independent and accepts no funding from AI companies or their staff.
Founders
Co-founder and CEO of Guidelight AI Standards. Over the past decade his work has focused on the risks and opportunities of advanced AI, including roles at OpenAI, Georgetown's Center for Security and Emerging Technology (CSET), and Longview Philanthropy. He holds a law degree from Cornell.
Co-founder and Chief Scientist of Guidelight AI Standards, where he leads the development of its standards and assessments. He previously worked at OpenAI (2020โ2024) on product safety and capability evaluations. He holds an MS in Computer Science from Georgia Tech and an AB in Economics from Brown.
Classification
External links
