Research Scientist - CAST Propensity

Core responsibilities:

• Study unprompted or unintended model behavior, particularly potentially dangerous behavior: the propensity of a model to cause harm
• Research effect sizes of environmental factors on these propensities, e.g., whether models are consistently more willing to take harmful actions when their existence is threatened
• Scale research to a range of different scenarios and variations, looking particularly for effects that are consistent throughout
• Help answer key research science questions through discussion, written plans and designs, and writing or reviewing code that implements those designs

This role creates positive change by providing empirical evidence with strong scientific credibility about AI model propensities, which is the key missing pillar in understanding risk from autonomous AI. This evidence helps justify responses sufficient to address unprecedented threats from advanced AI systems.

Required skills and qualifications:

• 3+ years of experience in a quantitative research discipline (e.g., as a PhD student or data scientist or researcher) involving experimental design and analysis
• Proven ability to identify and operationalize key uncertainties in a research area and propose/improve experimental approaches
• Knowledge and experience in selecting and applying statistical inference methods to draw risk-relevant conclusions from experimental evidence
• Strong Python knowledge to develop and iterate on Inspect tasks (Inspect can be learned on the job)
• Understanding of transformer architecture and training dynamics to inform interpretations and predictions of observable behavior

Compensation and benefits:

• Salary range: £65,000–£145,000 (base salary plus technical allowance)
• 28.97% employer pension contribution on base salary
• Modern central London office with cafes, food court, gym
• Hybrid working with flexibility for occasional remote work abroad
• At least 25 days' annual leave, 8 public holidays, extra team-wide breaks
• 3 days off for volunteering
• Generous paid parental leave (36 weeks of UK statutory leave shared between parents + 3 extra paid weeks)
• 5 days off learning and development, annual stipends for learning and development
• Funding for conferences and external collaborations
• Discounts and benefits for cycling to work, donations, and retail/gyms