Research Scientist, Manipulation Evaluations

• Design and build novel evaluations of AI harmful manipulation for EU AI Act enforcement.
• Develop risk scenarios then build and run evals end to end against frontier models.
• Design evaluation methods turning risk scenarios into rigorous, defensible measurements.
• Perform literature reviews, investigate model behaviour, and red-team your evals.

This role directly contributes to enforcing the EU AI Act by developing rigorous evaluations that detect and measure harmful manipulation by AI systems, thereby protecting society from deceptive and manipulative AI behaviors.

• Strong background in AI safety, machine learning, or related field.
• Experience designing and running evaluations or red-teaming of AI systems.
• Ability to write clearly for regulators and co-author technical reports.
• Familiarity with frontier models and manipulation risks.

Compensation and benefits not specified in the posting.

Apart Research is an organization focused on AI safety and governance, working to ensure that advanced AI systems are developed and deployed responsibly.