Anthropic Fellows Program 2027: Pay, Eligibility & How to Apply

Last updated: October 2026

The Anthropic Fellows Program is a paid, four-month, full-time research fellowship where you work on an empirical AI safety, security, systems or policy project with mentorship from Anthropic researchers. It pays $3,850 a week plus about $15,000 a month in compute. Applications are open for the January 2027 cohort: two tracks are rolling, and the AI Safety & Security track's deadline is October 30, 2026.

Quick Facts

The detail most applicants miss: the AI Safety & Security track is run through Constellation, Anthropic's recruiting partner, and you aren't considered unless you also submit Constellation's form. The Greenhouse application alone isn't enough for that track.

Anthropic says Python fluency and a strong technical background matter more than credentials. A remote Externship like the Hydroficient IoT Cyber Defense Externship gives you a finished technical project to point to in your application.

What Is the Anthropic Fellows Program?

The Anthropic Fellows Program is a funded research fellowship that Anthropic runs to grow AI research and engineering talent, "regardless of previous experience," in the words of its postings. Fellows spend four months full-time on an empirical project tied to Anthropic's research priorities, working mostly on external infrastructure like open-source models and public APIs. The goal is a public output, usually a paper submission. You're matched with a mentor and a project after you're accepted. There are three postings covering five research areas: AI safety (scalable oversight, adversarial robustness and AI control, model organisms, mechanistic interpretability, model welfare), AI security, ML systems and performance, reinforcement learning, and economics and policy through The Anthropic Institute. It isn't an internship and it isn't a full-time job; Anthropic's own degree policy for full-time hires explicitly doesn't apply to fellows. The program has had more than 200 fellows since early 2025.

Who Can Apply to the Anthropic Fellows Program?

Anthropic's Greenhouse postings set a short list of requirements:

• Python: you must be fluent in Python programming

• Availability: you must be able to work full-time on the fellowship for four months. If you can't commit to the full four months or the January start, say so in your application; Anthropic reviews those cases individually

• Background: a strong technical background in computer science, mathematics or physics is a good fit. The Economics & Policy track looks for interest in economics or policy research, with prior research experience not required

• Credentials: no PhD, prior ML experience or published papers needed, and no degree requirement is stated

• Work authorization: you must be authorized to work in the US, UK or Canada and be located in that country during the program. Anthropic doesn't sponsor visas for fellows

No rule excludes college students, but the fellowship is full-time for four months, so you'd need to take a term off or apply around graduation. The original 2024 program was aimed at mid-career technical professionals moving into AI safety research, and Anthropic still says it's open to applicants at any career stage.

Is the Anthropic Fellows Program Open for 2027?

Yes. As of October 30, 2026, all three Fellows postings are live on Anthropic's Greenhouse board for the cohort expected to start in January 2027. The ML Systems & Reinforcement Learning and Economics & Policy tracks review applications on a rolling basis with no deadline. The AI Safety & Security posting asks you to apply by October 30. Anthropic's 2026 announcement post gave an earlier date, October 18, but the job posting was updated on October 21 and is the one that counts. Anthropic runs several cohorts a year, so if you miss January, expect another intake later in 2027.

How to read this page: our dates come from tracking previous recruiting cycles and the employer's own postings. Treat future dates as informed estimates based on historical patterns, not confirmed dates. The program's official page is always the most current source. If it's posted there, go by that, even if this page suggests the window hasn't opened yet.

How Do You Apply to the Anthropic Fellows Program?

The steps depend on your track:

1. Pick your track. AI Safety & Security covers alignment, interpretability, model welfare and security research. ML Systems & RL covers performance engineering and reinforcement learning. The Anthropic Institute covers economics and policy research.

2. For AI Safety & Security, open the Greenhouse posting and complete Constellation's application form linked from it by October 30, 2026. The posting says you won't be considered without that form.

3. For the other two tracks, apply at the bottom of the ML Systems & RL or Economics & Policy posting. The form asks for your LinkedIn, website, publications and GitHub links, whether you've applied before, and a short "Why Anthropic?" answer.

4. Expect assessments and interviews. For Safety & Security, the process is an initial application and reference check, technical assessments and interviews, then a research discussion. The other tracks run an application, assessments and interviews, and may ask for references later.

5. Apply only through the official forms. Anthropic asks applicants not to reach out to mentors directly.

If you can't start in January or can't commit to the full four months, say so in your application rather than skipping it.

Does the Anthropic Fellows Program Lead to a Job at Anthropic?

Sometimes, but not by design. Anthropic's postings say 25% to 50% of fellows in previous cohorts received a full-time offer, and its 2026 announcement says more than 35% of past fellows went on to join Anthropic. Offers aren't guaranteed. The more reliable outcome is published research: over 65% of fellows have published work from their fellowship, and in one earlier cohort over 80% produced papers. Past fellow projects include research on subliminal learning in language models, open-source circuit tracing tools, and AI agents finding $4.6 million in smart contract exploits. For Anthropic's other early-career paths, see our Anthropic internship guide.

How Do You Stand Out for the Anthropic Fellows Program?

Anthropic doesn't ask for a PhD or published papers, so your application rests on what you've built and how clearly you think about research problems. A public GitHub with real projects does more than a list of courses. Here's how to build proof for each track:

Read a few papers from past fellows before your research discussion, and be ready to say which research area you'd pick and why.

What Similar Programs Should You Consider?

If you're looking at the Anthropic Fellows Program, these are worth a look too:

Outside the big labs, AI safety research programs like MATS and Constellation's own fellowships run on similar models. For the wider field, see our tech internships guide.

FAQ

How much does the Anthropic Fellows Program pay?

Fellows get a weekly stipend of $3,850 in the US (2,310 GBP in the UK or 4,300 CAD in Canada) plus benefits that vary by country. Anthropic also funds about $15,000 a month in compute and other research expenses. Over four months, the stipend comes to roughly $65,000.

What is the deadline for the January 2027 cohort?

It depends on the track. The AI Safety & Security posting asks you to apply by October 30, 2026. ML Systems & Reinforcement Learning and Economics & Policy review applications on a rolling basis with no published deadline. Anthropic runs several cohorts a year, so later intakes are expected.

Do you need a PhD to be an Anthropic Fellow?

No. Anthropic says you don't need a PhD, prior ML experience or published papers. The postings require Python fluency and full-time availability, and say a strong background in computer science, math or physics is a good fit. Anthropic's degree policy for full-time hires doesn't apply to fellows.

Is the Anthropic Fellows Program remote?

It can be. Fellows can work from shared workspaces in Berkeley and London for the Safety & Security track or San Francisco for the other tracks, or work remotely from the US, UK or Canada. You must be located in a country where you're authorized to work for the whole program.

Can international students apply?

Only if you already have, or independently get, full-time work authorization in the US, UK or Canada. Anthropic says it can't sponsor visas for fellows, and you must be located in the country where you're authorized during the program.

Can college students apply?

No rule excludes them, but the fellowship is full-time for four months, so you'd need to step away from classes or apply around graduation. Anthropic says it's open to applicants at any career stage and doesn't require a degree.

Does the Anthropic Fellows Program lead to a full-time job?

Often, but not guaranteed. Anthropic says 25% to 50% of fellows in previous cohorts received full-time offers, and more than 35% of past fellows went on to join the company. Over 65% of fellows have published work from their fellowship.

Who runs the application process?

For the AI Safety & Security track, applications and interviews are managed by Constellation, Anthropic's recruiting and program partner, which also runs the Berkeley workspace. You must complete Constellation's form to be considered. Anthropic handles the ML Systems, RL and Economics & Policy tracks directly.

Two tracks are rolling, and Anthropic runs more than one cohort a year, so a missed deadline isn't the end of it. If your GitHub doesn't yet show finished technical work, a remote Externship gives you a project to link before your next application.