Research Lead - AI Safety at FAR.AI

FAR.AI is a non-profit AI research institute focused on ensuring advanced AI is safe and beneficial for everyone. Since its founding in July 2022, FAR.AI has grown to 50+ staff, published 40+ academic papers, and convened leading AI safety events. Their work has been recognized globally, including a Best Paper Honorable Mention at ICML 2026, and features in prestigious publications like the Financial Times, Nature News, Wired Magazine, and MIT Technology Review. They conduct pre-deployment testing for frontier developers like OpenAI, evaluate governments including the EU AI Office, and publish the AI Security Leaderboard.

The role of Research Lead involves developing and leading a research agenda aimed at reducing catastrophic risks from advanced AI. You will:

• Define and own a research workstream end-to-end, including setting the research direction and mentoring a team of Members of Technical Staff.
• Build and lead a team executing the agenda, ensuring hands-on involvement in writing code and running experiments.
• Shape FAR.AI’s broader work by directing millions of dollars in grants, convening key stakeholders, and influencing independent testing and advising of AI companies and governments.
• Focus on empirically grounded, scalable ML safety work, prioritizing impact over novelty and intellectual elegance.

This role is ideal for someone who:

• Wants high autonomy in an impact-driven environment.
• Values collaboration with governments, leading AI companies, and academics.
• Has a strong research track record in AI or a highly technical field.
• Possesses a clear view of which safety research directions are likely to matter most over the next few years.
• Has experience leading teams, mentoring junior researchers, or supporting early-career researchers.
• Can effectively communicate novel methods to both technical and non-technical audiences.

FAR.AI is particularly interested in Research Leads in the following areas:

• Applied Interpretability — using interpretability to tackle concrete safety problems.
• Scalable Oversight / Alignment — methods to keep oversight robust as models become more capable.
• Adversarial Robustness — extending independent-testing work into deployed-system protection.
• Auditing / Evals — safety and alignment auditing, including evaluation awareness and black-box monitoring.
• Persuasion / Epistemic Risks — science of epistemic risks, persuasion's role in loss of control risks, and evaluations.
• Bring Your Own Agenda — open track for senior researchers with a strong vision outside the above pods.

Daily responsibilities include:

• Articulating a research agenda with a clear theory of change for mitigating catastrophic risks.
• Growing and leading a team of technical staff.
• Leading novel research projects with unclear markers of progress.
• Sharing research findings through written content and presentations.
• Mentoring and coaching junior team members.
• Contributing to the intellectual environment and giving feedback on early-stage proposals.
• Building a research field through grantmaking and events, and connecting it to real-world deployments.

This role is a great fit if you:

• Want to work on the most impactful research directions.
• Value empirically grounded, scalable research directions.
• Can speak freely and protect confidential information.
• Want to advise and collaborate with governments, leading AI companies, and academics.

This role is a poor fit if you:

• Prefer solo research to leading a team.
• Prioritize novelty and intellectual elegance over impact.
• Need the largest compute clusters or equity compensation.

To be a strong candidate, you likely:

• Have a strong research track record in AI or a related technical field.
• Have a clear view of which safety research directions are likely to matter most.
• Have either a clear research agenda or a strong track record with a developing agenda.
• Have led a team, mentored graduate students, or supported early-career researchers.
• Can effectively communicate novel methods to both technical and non-technical audiences.
• Are not a new entrant to AI safety.

It is preferable if you:

• Have an established publication record in AI safety.
• Are comfortable writing grant proposals and navigating collaborations.