Better evidence.
Built with industry.
Help shape independent research into how AI understands health, safety, and the environment.
Start a discovery conversationPractical questions. Independent evidence.
AI for HSE is an independent research programme led by Pavel Kosyrev. Our current benchmark, AI4HSE-Bench v1, measures model knowledge and answer reliability across health, safety, and environmental topics. It is a starting point for research, not evidence of readiness to make operational safety decisions.
Read the methodology or explore the published results.
Who we would like to hear from
Our Research Programme sets out the current benchmark and proposed directions for future studies.
Industrial operators and employers, HSE and occupational health teams, technology organisations applying AI to workplace decisions, and professional associations and research groups. We welcome organisations with practical questions about AI reliability and people with domain expertise to help investigate them.
Ways to contribute
- Research funding to support evaluations, analysis, and publication.
- Expert review of evaluation methods, question quality, and the interpretation of findings.
- Industry insight to identify relevant hazards, decision contexts, and questions for future research.
Research independence
Support cannot influence model rankings or publication decisions. AI for HSE retains control of evaluation methods, analysis, and what is published. Findings include failures and limitations. Any agreed research support and relevant conflicts of interest will be disclosed alongside the work.
Start with a conversation
Tell Pavel about your organisation, the questions you want to explore, and how you might contribute. A discovery conversation can help establish a useful research scope and suitable next steps. Partnership arrangements are exploratory; contributions, membership terms, and any benefits would be agreed together.