Head of AI Safety

Moonshot; Greater London; On-site; GBP 67,000 - 80,000; Full time

Benefits offered by this job

30 days' annual leave

Flexible public holiday policy

Private healthcare

Employee Assistance Programme

Maternity and paternity leave

Share options

Job summary

Moonshot in the United Kingdom is seeking a Head of AI Safety to lead the delivery, development, and growth of our AI Safety portfolio, combining violence prevention, behavioural risk, and online harms with evaluating AI system safety.

The role acts as Moonshot's primary safety counterpart for frontier AI companies, governments, and regulators, working with model, policy, trust and safety, product, research, and engineering teams to advance practical, ethical safeguards.

Qualifications

• Experience in trust & safety, online harms, or related field and ability to adapt to AI systems.
• Curiosity about AI with ability to engage credibly with technical teams.
• Experience designing research, evaluation frameworks, or interventions for harm categories.
• Experience managing projects, teams, budgets, partners, and clients.
• Excellent written communication for government, foundation, or enterprise audiences.
• Comfort handling highly sensitive or graphic content with wellbeing practices.
• Strong judgment navigating ambiguity and competing priorities.
• Willingness to travel and work outside regular hours when needed.
• Discretion and diplomacy, with readiness for security procedures.
• Experience supporting business development, grant funding, or procurement.
• Commitment to Moonshot's mission.

Responsibilities

• Lead and quality‑assure applied AI safety work across harm categories using red teaming and adversarial evaluation.
• Advise frontier AI companies on improving model safety, policies, and interventions.
• Translate insights from safety experts into actionable guidance for product, research, and engineering teams.
• Set the methodological approach for the portfolio and develop structured evaluation frameworks.
• Lead red teaming and adversarial evaluation with test scenarios and scoring criteria.
• Identify patterns and edge cases, and provide practical recommendations for model behaviour and user protections.
• Maintain rigorous documentation for technical, government, and foundation audiences.
• Ensure work complies with contractual, legal, data protection, and ethics obligations.
• Identify, manage, and escalate operational and partnership risks.

Skills

Trust & safety

AI safety curiosity

Research design

Project management

Written communication

Risk assessment

Ambiguity tolerance

Stakeholder management

Security clearance

Business development support

Mission alignment

Job description

Moonshot believes that marginalised people in society — including minority ethnic people, people from working class backgrounds, women, Disabled and LGBTQIA+ people — must be centred in the work we do. We strongly encourage applications from people with these identities or who are members of other communities who are currently underrepresented in our workforce. We know a diverse workforce will enable us to understand drivers behind violent extremism and online harms in an in-depth way and do better work to counter them.

About the role

Moonshot is recruiting a Head of AI Safety to lead the delivery, development, and growth of our AI Safety portfolio. The role combines Moonshot's expertise in violence prevention, behavioural risk, and online harms with the emerging practice of evaluating and improving the safety of AI systems. The portfolio addresses harm categories including pathways to violence, extremism, child sexual exploitation, abuse and grooming (CSEA), mental health and crisis, and risks affecting children and teenagers.

The Head of AI Safety will serve as Moonshot's primary applied AI safety counterpart for frontier AI companies, governments, and regulators. The role will work closely with model, policy, trust and safety, product, research, and engineering teams. This is not an engineering or data-science role, but it is a hands‑on position requiring the successful candidate to lead and participate directly in red teaming and adversarial evaluation, working in detail with evaluation methodologies, test scenarios, model responses, safety policies, and intervention frameworks. The role holds responsibility for client and partner relationships, project and staff management, methodological quality, and business development. The Head of AI Safety will build and maintain relationships across the wider AI safety ecosystem, including with governments, foundations, regulators, academics, researchers, and civil society organisations.

Your responsibilities will include:

Applied AI Safety, Evaluation, and Advisory

• Lead and quality‑assure Moonshot's applied AI safety work across harm categories including pathways to violence, extremism, CSEA, abuse and grooming, mental health and crisis, and risks affecting children and teens, using methods such as red teaming and adversarial evaluation of AI systems.
• Advise frontier AI companies on how to improve the safety of their models, products, policies, and intervention systems.
• Translate insights from psychologists, child‑safety specialists, violence‑prevention practitioners, safeguarding experts, and other subject‑matter experts into clear, actionable guidance for model safety, policy, product, research, and engineering teams.
• Set the methodological approach for the portfolio, translating violence‑prevention, safeguarding, and behavioural‑risk expertise into structured and testable evaluation frameworks.
• Lead and participate directly in red teaming and adversarial evaluation, working in detail with test scenarios, model responses, scoring criteria, safety policies, and evaluation results.
• Identify patterns, edge cases, and potential safety failures, and develop practical recommendations for improving model behaviour and user protections.
• Maintain rigour and clear documentation across the team's technical deliverables, suitable for technical, government, and foundation audiences.
• Ensure work is delivered within a clear ethical framework and in compliance with contractual, legal, data protection, and ethics obligations.
• Identify, manage, and escalation operational, reputational, delivery, and partnership risks.

Client & Partner Management

• Serve as Moonshot's primary applied AI safety counterpart for frontier AI company partners, governments, regulators, and the wider ecosystem invested in AI safety.
• Build trusted relationships with model, policy, trust and safety, product, research, and engineering teams.
• Build and sustain relationships across the wider AI safety ecosystem, including governments, foundations, regulators, academics, researchers, civil society organisations, and specialist practitioners.
• Represent Moonshot externally in meetings, briefings, workshops, and sector engagement, including with regulators and policymaker audiences.

Team Leadership & Management

• Provide direct leadership, coaching, and management to Moonshot's AI safety team.
• Foster a collaborative, accountable, and mission‑driven team culture, with particular attention to wellbeing given the sensitive nature of the work.
• Support workforce planning, performance management, and professional development across the team.
• Ensure effective coordination with internal teams supporting the portfolio, including operations, finance, research, and technical teams.

Portfolio Development & Growth

• Develop Moonshot's AI safety portfolio, identifying strategic opportunities, partnerships, and funding.
• Lead proposal development, scoping, and renewals with technical credibility, using precise, defensible language suited to technical and government audiences.
• Develop repeatable methodologies, service offerings, and partnerships that allow the portfolio to grow while maintaining methodological rigour and delivery quality.
• Support external communications, publications, briefings, and thought leadership that establish Moonshot as a credible voice in applied AI safety.
• Oversee project planning, staffing, budgeting, forecasting, and delivery timelines across the portfolio.

Essential:

• Experience in trust & safety, online harms, or a closely related field such as violence prevention, safeguarding, or public health, and the ability to adapt that knowledge to AI systems.
• Curiosity about AI and the ability to build technical fluency quickly, enough to engage credibly with technical counterparts at AI companies. Much of this work is new, so comfort learning as you go matters more than existing AI safety expertise.
• Experience designing research, evaluation frameworks, or interventions for harm categories such as violent extremism, CSEA, self‑harm and crisis, or targeted violence.
• Demonstrated experience managing projects, teams, budgets, partners, and clients, with strong people management skills.
• Excellent written communication, with experience producing credible (not promotional) material for government, foundation, or enterprise audiences.
• Comfort and demonstrated resilience working with highly sensitive or graphic content (CSEA, extremist material, crisis content), with awareness of wellbeing practices for this kind of work.
• Strong judgment and the ability to navigate ambiguity, competing priorities, and sensitive stakeholder environments, including representing organisations externally.
• Willingness to travel and work outside regular hours where needed to accommodate clients or respond to incidents.
• Highly trustworthy, with discretion and diplomacy, and willing to undertake relevant security clearance procedures.
• Experience supporting business development, grant funding, or procurement.
• Commitment to Moonshot's mission.
• We require and will check candidates' eligibility to work in the UK and pass any relevant security clearance procedures per the needs of clients.

Desirable:

• Direct experience in model safety, red teaming, or adversarial evaluation of LLMs or other AI systems.
• Understanding of LLM architecture, safety tooling, or trust & safety policy.
• Prior experience in child safety evaluation, teen‑safety product work, or grooming and CSEA detection.
• Familiarity with government or regulatory engagement, such as briefing officials or supporting policy submissions.
• Experience with intervention or diversion programme design that can transfer to AI-mediated interventions.
• Academic or applied background in radicalisation studies, forensic psychology, or violence risk assessment.
• Familiarity with taxonomy or classifier development, including how testing data feeds a classifier.

• 30 days' paid annual leave, excluding bank holidays.
• Flexible public holiday policy with the option to work bank holidays in exchange for a day off at another time.
• Private healthcare package with access to specialist mental health cover, including coverage for partners and children.
• Employee Assistance Programme providing access to mental health support.
• Generous maternity and paternity leave: 26 weeks paid maternity leave, 8 weeks paid paternity leave.
• All permanent employees are granted share options upon employment.

Salary: £67,000 - £80,000 (depending on skills and experience).