Build before the flood
Study AI risks before a disaster forces the lesson.
An Amsterdam research community making advanced AI safe, secure and well governed.
Five principles, borrowed from how the Dutch live with water.
Study AI risks before a disaster forces the lesson.
Layered safeguards across design, security, evaluation, deployment and law.
Safety is kept by many people, not owned by one lab or government.
Technical research connected to law and policy, including the EU AI Act.
We welcome disagreement and work toward common ground.
AI safety, security and governance research.
Short papers, evaluations and open-source tools.
Testing AI systems for misuse and failure.
Technical findings for regulators and organizations.
Notes, code and results at dijkai.github.io.
Researchers, engineers, lawyers, students, policymakers. No expertise or Dutch needed; we work in English.