About the role
ABOUT US
FAR.AI http://FAR.AI is a non-profit AI research institute working to ensure advanced AI is safe and beneficial for everyone. Our mission is to facilitate breakthrough AI safety research, advance global understanding of AI risks and solutions, and foster a coordinated global response.
Since our founding in July 2022, we've grown to 50+ staff https://www.far.ai/about/team, published 40+ academic papers https://scholar.google.com/citations?user=FVJ24k8AAAAJ, and convened leading AI safety events https://far.ai/events/. Our work is recognized globally, with publications at premier venues such as NeurIPS, ICML including a Best Paper Honorable Mention in 2026 https://icml.cc/virtual/2026/oral/71065, and ICLR, and features in the Financial Times https://www.ft.com/content/175e5314-a7f7-4741-a786-273219f433a1, Nature News https://www.nature.com/articles/d41586-024-02218-7, Wired Magazine https://www.wired.com/story/jailbreaking-ai-models-google-anthropic-openai-spacexai/ and MIT Technology Review https://www.technologyreview.com/2020/02/28/905615/reinforcement-learning-adversarial-attack-gaming-ai-deepmind-alphazero-selfdriving-cars/. We conduct pre-deployment testing on behalf of frontier developers such as OpenAI, independent evaluations for governments including the EU AI Office https://www.far.ai/news/far-ai-selected-to-lead-eu-ai-act-cbrn-risk-consortium, and publish the AI Security Leaderboard https://leaderboard.far.ai/ based on our red-teaming expertise. We help steer and grow the AI safety field through developing https://arxiv.org/abs/2405.06624 research https://arxiv.org/abs/2506.20702 roadmaps https://www.researchgate.net/publication/396910034_Open_Technical_Problems_in_Open-Weight_AI_Model_Risk_Management with renowned researchers such as Yoshua Bengio; running FAR.Labs https://www.far.ai/programs/far-labs, an AI safety-focused co-working space in Berkeley housing 40+ members; and supporting the community through targeted grants https://www.far.ai/programs/grantmaking to technical researchers.
THE ROLE
FAR.AI http://FAR.AI is seeking a Loss of Control Lead to create and deploy evaluations that uncover and lead to solutions for risks of autonomy and RSI. You will lead development of a testing program in this area, combining rigorous threat modeling and evaluation science, applied testing, and building an elite team to do it. Your program will work closely with frontier AI companies and leading governments, producing independent evaluation insights that greatly accelerate the world’s progress on critical safety outcomes.
As Loss of Control Lead in the FAR.AI http://FAR.AI Frontier Safety division, you will own evaluation across a risk domain, reporting to Kellin Pelrine with a dotted line to Edward Yee. You will partner closely with Kellin, the division’s technical lead, to build your testing program: you and the team you build will lead what to test and how to do it, while Frontier Safety and FAR.AI http://FAR.AI as a whole will support with frontier lab access, engineering and infrastructure, recruiting, and other systems that clear your path to impact.
There is also an individual contributor (IC) track for this role should you wish to not manage or build a team. We are open to several alternative versions of this role, including both senior IC versions that emphasize hands-on work without management duties, and manager-of-managers versions for senior leaders aiming to build multiple teams. Simply apply or reach out to talent@far.ai.
WHY THIS ROLE / WHY NOW
FAR.AI http://FAR.AI’s Frontier Safety division exists to materially raise the bar for safety and security of the most capable AI systems in the world. We intend to be the tip of the spear in AI safety: the team that consistently finds the failures others miss, converts these into mitigations by frontier developers, and sets the standard that labs and governments converge on.
The division grew out of FAR.AI http://FAR.AI's red team, and is expanding rapidly into loss of control testing, deeper evaluations, and safeguards. From that base, we discover issues, drive their mitigation, and, where a gap in safeguards is blocking a fix, build the missing technical capacity ourselves. What we learn from building and studying safeguards then feeds back into new ways of finding vulnerabilities.
We are already one of the leading independent red-teaming organizations globally. Our work has helped most Western frontier model developers improve safeguards through pre- and post-deployment testing (e.g., OpenAI, Anthropic, and Google) and pioneered new industry standards (e.g. the FAR.AI http://FAR.AI Minimal Standard for Safeguards). We are also embedded in high-leverage government efforts (e.g., leading a consortium building CBRN evaluations for the European Commission/EU AI Office, and collaborating with the UK AI Security Institute).
Success here looks like control failures identified, safeguards strengthened, and global standards shifted – working with the leading frontier labs and informing the public, scientists, and governments globally to make this happen. We expect this role to hit the ground running, deploying testing through our established lab partners in the first month, and beginning hiring and teambuilding in the first quarter.
KEY RESPONSIBILITIES
Build a program of high impact loss of control testing with frontier AI companies and governments
- Build evaluations and other testing and analysis tools that decisively uncover bad behavior and safeguard gaps in frontier AI systems
- Develop rigorous and systematic testing methodologies and playbooks that produce credible, decision-relevant evidence about loss of control risks, including illuminating the conditions under which failures occur, evaluating the effectiveness of mitigations, and identifying important limitations in what current evaluation methods can establish
- Strategize and execute an impact-maximizing evaluation portfolio, across a range of settings like early in model training to post deployment, and simple inference-only access to deep embedded evaluation
- Ship private reports to key stakeholders like frontier companies and governments, informing and enabling action to improve safeguards
- Create and support public reports, benchmarks, and leaderboards that shape industry norms
- Fill holes in threat modeling and risk analysis, creating the end-to-end understanding and saliency needed to solve safety challenges
Manage, mentor, support, and grow an elite team
- Plan sprints and lead execution with extremely high standards for velocity, rigor, judgment, and impact
- Mentor team members in scientific and other skills, both in direct 1-1 settings and by building resources for rapid skill growth and knowledge transfer
- Manage team members, supporting and amplifying their impact, setting high standards and prioritization week-to-week, and creating the environment and paths for rapid growth
- Create processes that scale high-tempo innovation without sacrificing quality
Collaborate across FAR.AI http://FAR.AI and the field
- Collaborate with leads in other areas to improve internal practices for rigorous, scalable, and impact-maximizing evaluation, and disseminate key findings to lift the public independent testing ecosystem
- Contribute to the overall research and testing strategy of the Frontier Safety division
- Partner closely with the rest of Frontier Safety to translate technical ideas and findings into real-world impact
IS THIS ROLE A FIT?
This role would be a great fit if you
- Are excited by high-stakes, real-world technical work where success is measured by impact, not revenue, clicks, or papers published
- Want to work with a range of leading frontier AI companies, researchers, and governments. We’re a lean organization, and seek to leverage our impact through strategic partnerships
- Get energy from working at the frontier of AI and shaping its path forward
-