AI Standards Lab
An independent nonprofit and affiliated research company dedicated to accelerating the development of AI safety standards and risk management frameworks, with a focus on EU AI Act standards and global AI safety engineering.
An independent nonprofit and affiliated research company dedicated to accelerating the development of AI safety standards and risk management frameworks, with a focus on EU AI Act standards and global AI safety engineering.
People
Updated 05/18/26Co-founder and board president
Co-founder and board member
Co-lead and resident standards expert
Co-lead and research analyst
Research analyst
Researcher
Legal and policy researcher (EU AI Act standards)
Researcher
Funding Details
Updated 05/18/26- Annual Budget
- -
- Current Runway
- -
- Funding Goal
- -
- Funding Raised to Date
- $528,000
Org Details
Updated 05/18/26AI Standards Lab and Holtman Systems Research are two closely linked organizations working to accelerate the development of AI safety standards and risk management frameworks worldwide. The AI Standards Lab was founded in November 2023, growing out of a volunteer AI Safety Camp project that ran from March to June 2023. It was incorporated as a US nonprofit in April 2024 and operates as a Delaware nonprofit corporation through fiscal sponsorship with Players Philanthropy Fund, Inc. The Lab functions as a virtual organization, bringing together experts from around the world with backgrounds in computer science, AI research, safety engineering, standards development, and law. Holtman Systems Research is a small European company founded in November 2022 by Koen Holtman, a systems architect and independent AI safety researcher based in Eindhoven, the Netherlands. Holtman holds a PhD in Software Design from Eindhoven University of Technology, earned through research conducted at CERN. He has 20 years of experience in industrial research and development and 10 years in standards creation, with contributions to the HTTP/1.1 protocol and Blu-ray Disc system standards. His AI safety research includes published work on corrigibility and agent foundations. The organizations share three main focus areas. First, they support EU AI Act implementation by contributing to technical standards development through the CEN-CENELEC JTC21 committee and the General-Purpose AI Code of Practice, emphasizing clear minimum safety requirements. Second, they work on global AI safety engineering, developing standardized approaches for evaluating frontier models and analyzing control loss risks. Third, they produce research and publications including reports on AI Act amendments, a Quality Scorecard for AI Evaluations, and an extensive catalog of risk sources and risk management measures for general-purpose AI systems. The AI Standards Lab team includes co-leads Koen Holtman and Ze Shen Chin (a research affiliate at the Oxford Martin AI Governance Initiative), research analysts Adrian Regenfuss, Ayrton San Joaquin, Marcel Mir Teijeiro, and Rokas Gipiskis, and operations specialist Yi-Yang Chua. The board includes co-founders Ariel Gil (Board President, Technical Research Manager at Pivotal) and Jonathan Happel (CEO at TamperSec). Recent public outputs include analysis of European Parliament amendments to the EU AI Act in the Digital Omnibus (March 2026), a Quality Scorecard framework for AI evaluations (February 2026), feedback on transparency codes of practice for AI-generated content (January 2026), and input to the European Commission on the future of European standardisation (December 2025).
Theory of Change
Updated 05/18/26AI Standards Lab and Holtman Systems Research believe that AI safety standards must keep pace with the rapid development and deployment of AI technology, and that well-designed technical standards are a critical mechanism for ensuring safe AI outcomes. Their theory of change centers on three pathways: first, by directly contributing expert knowledge to formal standards bodies like CEN-CENELEC JTC21, they help shape the mandatory technical requirements that AI developers must meet under the EU AI Act, establishing minimum safety floors that prevent competitive pressures from degrading safety. Second, by developing frameworks like the Quality Scorecard for AI Evaluations and cataloging risk sources for general-purpose AI, they provide the analytical groundwork that standards bodies need to write effective, technically grounded standards. Third, by operating as an independent virtual lab that brings together diverse experts, they help bridge the gap between AI safety research and the standards-making process, ensuring that safety-relevant technical knowledge is translated into enforceable requirements.
Grants Received
Updated 05/18/26Projects– no linked projects
Updated 05/18/26Discussion
Key risk: Their standards-first theory of change risks dilution and harmful lock-in—producing slow, weak, or mis-specified requirements for frontier models—and as a small early-stage team their marginal influence may be limited or counterfactually replaceable by incumbent standards bodies.
Case for funding: With a track record of getting their work officially incorporated into EU AI standards (via CEN-CENELEC JTC21 and the General-Purpose AI Code of Practice), AI Standards Lab can translate safety research into enforceable minimum requirements through concrete artifacts like their Quality Scorecard and risk catalog—a high-leverage bottleneck in AI Act implementation that few orgs are positioned to fill.