We're actively hiring across research, engineering, policy, and operations, and we offer fellowships for early-career researchers and practitioners. If you don't see a perfect fit but believe you can contribute to our mission, reach out anyway.
Research, engineering, and adversarial testing roles for building safer AI systems.
Conduct cutting-edge research on AI safety, focusing on adversarial robustness, interpretability, and alignment. Publish findings and contribute to safety tools.
Build and maintain secure infrastructure for AI model evaluation and deployment. Design security controls, implement monitoring systems, and respond to incidents.
Lead adversarial testing campaigns against frontier AI models. Develop novel jailbreak techniques, identify security vulnerabilities, and assess misuse potential.
Build evaluation frameworks, benchmarks, and tools for AI safety research. Collaborate with researchers to translate ideas into production-ready systems.
Policy roles translating AI safety research into practical guidance and standards.
Shape AI safety policy, contribute to standards development, and engage with government agencies. Translate technical research into actionable policy recommendations.
Flexible, stipended placements for early-career researchers and practitioners.
A fellowship for early-career researchers and industry practitioners in AI security and related fields. Fellows contribute to our research on AI-enabled scams and social engineering. Exceptional proposals for independent work are also considered.
A fellowship for early-career researchers and practitioners in technology policy, law, or AI governance. Fellows translate our research on AI-enabled scams into policy analysis, standards input, and briefings for regulators and industry. Exceptional proposals for independent work are also considered.
We're always interested in hearing from talented people who care about AI safety. Whether you're interested in full-time, part-time, or advisory roles—reach out and tell us how you'd like to contribute.