How to apply for Anthropic Fellows Program, AI Safety & Security

About Anthropic

Anthropic is a frontier AI research and product company working on alignment, policy, and security. It posts specific high-impact opportunities, and it explicitly notes it does not necessarily recommend working at other positions there. Candidates should read Anthropic's own career review about concerns of doing harm by working at a frontier AI company.

About the role

The Anthropic Fellows Program funds and mentors promising technical talent, regardless of previous experience, to work on an empirical project aligned with Anthropic's research priorities. Fellows work full-time for 4 months using external infrastructure such as open-source models and public APIs, with the goal of producing a public output like a paper submission. In one earlier cohort, over 80% of fellows produced papers.

A typical day

A fellow spends most of the day on their empirical project, using open-source models or public APIs and a compute budget of about $15k/month. They meet with an Anthropic researcher mentor and use a shared workspace in Berkeley or London, or work remotely if US-based. The job description does not specify a daily routine beyond this, so ask Anthropic about meeting cadence, workspace expectations, and how remote fellows collaborate.

Who Anthropic is looking for

• Has strong technical skills in AI research or engineering, even without prior experience at a frontier lab.
• Can design and run an empirical project using external infrastructure such as open-source models or public APIs.
• Has a clear interest in AI safety and security research, and can connect their project to Anthropic's research priorities.
• Can commit to 4 months of full-time research and produce a public output, such as a paper submission.

Tips for this application

• Apply by October 18 for the January cohort. Applications are reviewed on a rolling basis, so earlier is better.
• Propose a concrete empirical project that can be run with open-source models or public APIs, since fellows primarily use external infrastructure.
• State how your project aligns with Anthropic's research priorities in AI safety and security, not just general AI interest.
• Show that you can produce a public output, such as a paper submission, within 4 months of full-time work.
• Address the 80,000 Hours career review concern about working at a frontier AI company directly if it is relevant to your motivation.

What to cover in your cover letter

["A specific empirical research project you want to pursue during the fellowship and why it fits Anthropic's safety and security priorities.", 'Evidence you can execute full-time research for 4 months and produce a public output, such as a prior paper, report, or code release.', 'Your technical ability to work with open-source models and public APIs, since fellows use external infrastructure.', "Your fit for the program's goal of fostering AI research and engineering talent, including any non-traditional path into the field."]

Research before applying

• Read Anthropic's mission statement and its stated focus on reliable, interpretable, and steerable AI systems.
• Read the 80,000 Hours career review linked in the job description about concerns of doing harm by working at a frontier AI company.
• Review Anthropic's published research priorities in alignment, policy, and security to identify a project that fits.
• Check the logistics: the shared workspace is in Berkeley, California or London, UK, and the role is listed as remote in the US, so clarify location expectations with Anthropic.

Likely interview topics

Based on the job description, expect questions about:

• Your proposed empirical project: feasibility, methodology, compute needs, and expected public output.
• How your project aligns with Anthropic's mission to create reliable, interpretable, and steerable AI systems.
• Your experience using open-source models or public APIs for research, and how you would handle external infrastructure limits.
• How you would use the ~$15k/month compute budget and mentorship from Anthropic researchers.
• Your understanding of AI safety and security research priorities, and how you would connect to the broader community.

Common mistakes to avoid

• Proposing a project that requires infrastructure or data only available inside a frontier lab, since fellows primarily use external infrastructure.
• Applying without a concrete plan for a public output, such as a paper submission, within the 4-month fellowship.
• Treating the fellowship as a general AI job application without connecting your work to AI safety and security research priorities.