Governance Researcher (Expression of Interest)

ABOUT APOLLO RESEARCH

Apollo Research is a mission-driven startup committed to reducing the threats from scheming frontier AI systems. We work with and are trusted by every frontier AI lab. We do fundamental research, test their models before deployment, and collaborate with them on scheming mitigations.

ABOUT THE TEAM

The Governance team conducts research and provides advice on the institutional, legal and strategic implications of AI scheming and loss of control threats, working closely alongside our Science of Scheming, Evaluations, and Monitoring teams. The Governance team works on a bet: that the governance, risk management and legal frameworks developed and implemented within and outside frontier labs in the next two years will contribute to determining our collective ability to stay in control of increasingly advanced AI systems. That means we focus on researching some of the hardest open problems in the field: scheming, loss of control, the governance of models deployed internally by frontier labs, and the imminent handoff of AI research and development to AIs themselves. Our outputs include the loss of control chapter of the International AI Safety Report 2026, the landmark report AI Behind Closed Doors: A Primer on the Governance of Internal Deployment, The Loss of Control Playbook:Degrees, Dynamics, and Preparedness, a TIME op-ed with Yoshua Bengio, peer-reviewed work on precursory capabilities, AI incident regimes and National Security Benchmarks for Loss of Control presented at ICML, IASEAI and the UK AISI’s Conference on Frontier AI Safety Frameworks. We are routinely requested to brief members of governments on both sides of the Atlantic, including members of Congressional committees, the intelligence community and offices of Members of Congress. A portion of our work is non-public, written for decision makers across key nodes.

The work. Our research team is empowered to conduct independent, high-quality research on the most consequential open questions raised by advanced AI and associated threats, shaping the conversation rather than executing against existing presets. Our current research priorities are threat modelling and mitigations for loss of control, the governance of internal deployment, and developing safe step changes towards automated AI R&D (with a particular interest in AI handoff). You can learn more about our work here and here.