Workstream Lead, Agentic AI Risk Modelling and Mitigations

Written by the T&S Network from AI Security Institute's posting.

Leads a government research team that builds scenario-based risk models on how advanced AI systems could become hard to oversee, correct or shut down as they take on longer, more autonomous tasks. The team identifies mitigations such as monitoring infrastructure, deployment guidance and containment protocols, and works to get them adopted by UK organisations best placed to act.

What you'd do

• Set the team's strategy, quarterly plan and overall research agenda
• Choose the analytical approaches and which risk scenarios the team develops
• Review the quality of analyses and outputs before they go out
• Convert the team's findings into assessments, mitigation proposals and routes to implementation
• Work with UK organisations to support adoption of existing mitigations
• Recruit, manage and grow a mixed-discipline team running several projects at once
• Build relationships with senior policy and technical staff in government, research bodies and frontier AI companies

What it asks for

• Experience leading impactful research, scientific work or technical programmes in frontier AI, AI safety, AI governance, risk analysis or a closely related field
• Past work planning and completing research programmes in technically intricate domains
• Proven success managing strong mixed-discipline teams and mentoring less senior colleagues
• Past work turning technical results into actionable advice, operational changes or policy material
• Experience building relationships with senior stakeholders in government, academia or industry
• Strong written and verbal communication for technical and non-technical audiences
• Knowledge of AI control, alignment, model autonomy, scenario work, systemic risk or frontier AI governance (preferred)

Worth knowing

• Successful candidates need a criminal record check and BPSS clearance; CTC eligibility is strongly preferred

Our best summary of the posting, and it can be wrong. The employer's posting is what counts: read it in full on AI Security Institutenorth_east before you apply. If we got something wrong, report this listing.