Agentic AI Risk Modelling and Mitigations

Core responsibilities:

• Develop rigorous models of how agentic AI could cause harm, identifying practical mitigations with a focus on measures the UK government are well-placed to implement.
• Draw on empirical evidence from AISI's evaluations and broader cybersecurity/ML literature to develop detailed threat models and mitigations.
• Reason carefully about complex and uncertain scenarios and communicate findings clearly to both technical researchers and policy decision-makers.
• Collaborate with government partners on hands-on ML or cybersecurity work to develop mitigations.

This role creates positive change by developing risk models and mitigations for agentic AI systems to prevent loss of control scenarios, directly influencing how frontier AI is governed and deployed globally through work with the UK government and leading AI companies.

Required skills and experience:

• Experience producing detailed threat models, risk analyses, safety cases, or similar structured analytical work in AI safety, cybersecurity, national security, or another domain.
• A track record of published research or substantial written analysis demonstrating rigorous reasoning about complex, uncertain topics.
• Strong written communication: ability to present complex technical arguments clearly to both technical and non-technical audiences.
• Deep familiarity with cybersecurity and how it will be impacted by high-capability AI agents, or with AI alignment and safety literature including loss of control, deception, power-seeking, scalable oversight, and AI control.
• Sense of mission, urgency, and responsibility for success; ability to work self-directed while collaborating effectively.

Compensation and benefits:

• Salary ranges from £65,000 to £145,000 total package depending on level (Level 3: £65,000-£75,000, Level 4: £85,000-£95,000, Level 5: £105,000-£115,000, Level 6: £125,000-£135,000, Level 7: £145,000)
• 28.97% pension contribution on top of base salary
• At least 25 days annual leave, 8 public holidays, extra team-wide breaks, 3 days off for volunteering
• Generous paid parental leave (36 weeks UK statutory leave shared between parents + 3 extra paid weeks)
• 5 days off learning and development, annual stipends for learning, funding for conferences
• Hybrid working with modern central London office or option to work in other UK government offices
• Flexibility for occasional remote work abroad, stipends for work-from-home equipment
• Pre-release access to multiple frontier models and ample compute resources
• Discounts for cycling to work, donations, retail/gyms

The AI Security Institute is the world's largest and best-funded team dedicated to understanding advanced AI risks and translating that knowledge into action, working within the UK government with direct lines to the Prime Minister's office and collaborating with frontier developers and governments globally.