L’essentiel de l’offre
- Poste : Anthropic Fellows Program in AI Safety and Security
- Employeur : Anthropic
- Lieu : London, United Kingdom; Berkeley, California; and eligible remote locations in the US, UK, and Canada
- Contrat : Four-month programme, full-time
- Date limite : 18 October 2026
- Niveau requis : Strong technical background in computer science, mathematics, physics, or another relevant discipline
Contexte
Anthropic has opened applications for its January cohort of the Anthropic Fellows Program in AI Safety and Security, offering aspiring researchers and engineers an opportunity to undertake intensive AI research with expert mentorship, financial support, and research funding. The four-month programme is designed to develop technical talent committed to building reliable, interpretable, and steerable artificial intelligence systems.
The fellowship welcomes promising technical candidates, including those seeking to transition into empirical AI research. Participants will work on research projects aligned with Anthropic’s priorities, with the goal of producing public outputs such as academic papers and other research contributions.
About the Anthropic Fellows Program: Anthropic is an artificial intelligence company focused on developing AI systems that are safe, beneficial, understandable, and controllable. Its Fellows Program supports researchers and engineers through financial assistance, direct mentorship, and opportunities to contribute to important technical questions in AI safety and security.
Fellows primarily use external research infrastructure, including open-source models and public application programming interfaces, to conduct empirical research. Anthropic reports that more than 80% of fellows in one previous cohort produced papers.
The programme also connects participants with the wider AI safety and security research community, potentially helping them establish research collaborations and pursue future opportunities in the field.
Missions et responsabilités
- Work on research projects aligned with Anthropic’s priorities, with the goal of producing public outputs such as academic papers and other research contributions.
- Focus on the AI Safety Fellows track: reducing risks associated with increasingly capable artificial intelligence systems, including scalable oversight, adversarial robustness and AI control, model organisms, mechanistic interpretability, and AI welfare.
- Focus on the AI Security Fellows track: concentrating on identifying and addressing vulnerabilities in AI systems and related technologies, including penetration testing, vulnerability research, security evaluations, and the development of techniques for assessing advanced AI systems.
- Dedicate approximately 40 hours of work per week to the full-time research programme.
Profil recherché
- Strong technical background in computer science, mathematics, physics, or another relevant discipline.
- Complementary expertise in economics, social sciences, cybersecurity, or related research areas for specific workstreams.
- Demonstrate fluency in Python programming.
- Be available for full-time participation.
- Have full-time work authorisation in the United States, United Kingdom, or Canada, and be able to reside in their authorised country throughout the programme.
- Demonstrate initiative, technical problem-solving skills, and the ability to make progress on complex, ambiguous research problems.
- Experience in empirical machine learning research, large language models, open-source contributions, or related technical disciplines for the AI Safety track.
- Experience contributing to open-source repositories related to large language models or cybersecurity, reporting common vulnerabilities and exposures (CVEs), earning bug bounties, or working with deep learning frameworks and experiment management tools for the AI Security track.
Conditions et avantages
- Weekly stipends of US$3,850, £2,310, or CAD$4,300, depending on the participant’s country.
- Direct mentorship from experienced Anthropic researchers.
- Research computing funding of approximately US$15,000 per month, alongside funding for other approved research expenses.
- Access to shared workspaces in Berkeley, California, or London, United Kingdom.
- Opportunities to engage with the broader AI safety and security community.
- Potential pathways towards future research and engineering positions (25–50% of fellows in previous cohorts received full-time offers, though employment is not guaranteed).
Dossier de candidature
- Updated CV highlighting Python proficiency, technical projects, research experience, and relevant open-source contributions.
- Required application completed through the official Constellation application form, including explanations of individual contributions to previous projects, measurable technical achievements, and links to relevant publications, code repositories, or research outputs.
- Circumstances explanation for candidates unable to commit to the entire four-month period.
Comment postuler
Applications and interviews are managed by Constellation, Anthropic’s recruiting and programme partner.
- Review the AI safety and security research areas to identify the most suitable track.
- Prepare an updated CV highlighting Python proficiency, technical projects, research experience, and relevant open-source contributions.
- Submit the required application through the official Constellation application form. (Important: Applicants must complete the Constellation application form to be considered. The separate Greenhouse application is not required.)
- Prepare for reference checks, technical assessments, interviews, and a research discussion.
Date limite : 18 October 2026
VISIT OFFICIAL WEBSITE TO APPLY