Anthropic Internship 2027–2028: Programs, Deadlines & How to Apply

Last updated: July 2026

Anthropic is valued at $965 billion after its Series H in May 2026, making it the most valuable pure-play AI company on Earth.

But here's the thing most applicants miss: Anthropic doesn't offer a traditional internship. Instead, the Anthropic Fellows Program is a 4-month paid research fellowship that pays $3,850 per week, provides $15,000 per month in compute credits, and converts over 40% of participants to full-time roles. For the 2027-2028 cycle, cohorts are expected to start quarterly (January, May, July, September), and applications are reviewed on a rolling basis until spots fill.

So the question isn't whether you're qualified on paper. It's whether you can show genuine AI safety motivation and practical engineering ability before the next cohort closes.

Quick Facts

The numbers that matter: rolling applications reviewed continuously, a 4-month fellowship paying $3,850/week plus $15,000/month in compute, and a 40%+ conversion rate to full-time roles at a $965B company. Anthropic does not sponsor visas for fellows.

Externships are short, remote professional experience programs where you finish a real project with a real company. The Wayfair AI Agent Engineering Externship and the Beats by Dre Data Analytics Externship build exactly the Python-based ML engineering and data analysis evidence that Anthropic's rolling fellowship applications reward. Explore all Externships.

What Is an Anthropic Internship?

An Anthropic fellowship isn't a traditional internship. It's a 4-month paid research residency where you work on live AI safety problems alongside the team that built Claude. The Anthropic Fellows Program places you in one of five research workstreams (AI Safety, AI Security, Reinforcement Learning, ML Systems, or Mechanistic Interpretability) and gives you the autonomy to drive your own project from day one.

And the results speak for themselves: over 80% of fellows produce published research papers during their tenure. With $47 billion in annual recurring revenue and 2,300+ employees working across 12 global offices, Anthropic operates at a scale where your fellowship project contributes to production systems rather than sitting in a sandbox. The culture is mission-driven and intellectually intense, rooted in the founders' conviction that building safe AI requires bringing in people who can think clearly about hard technical questions and write production-quality code under ambiguity.

When Do Anthropic Internship Applications Open for 2027–2028?

Anthropic runs multiple fellowship cohorts per year on a rolling calendar. Based on prior cycles, the 2026 cohorts started in May, July, and September/October, with applications opening 2 to 3 months before each start date. For 2027, expect quarterly cohort starts in January, May, July, and September. Because applications are reviewed continuously until spots fill, the proof you bring to your application needs to be built before the next window opens.

And unlike batch-processed programs at larger companies, Anthropic moves fast: the entire process from application to offer takes approximately 3 to 6 weeks.

Why You Must Apply the Week Applications Open

Anthropic reviews applications on a rolling basis, and the data confirms what that means in practice. Each cohort has approximately 32 spots, and roughly 150 candidates make it through all interview rounds for those seats. That's already a 21% final-round acceptance rate.

But the real filter happens earlier: the CodeSignal assessment eliminates 60 to 70% of applicants before they ever speak to a human.

So why does timing matter? Because rolling review means early applications get processed first, and cohort spots fill as offers are accepted. Waiting until the listing has been up for weeks means competing for fewer remaining seats.

And there's no stated deadline to anchor against because Anthropic simply closes the listing when the cohort is full.

Which Anthropic Internship Programs Should You Target?

Anthropic doesn't run traditional internship tracks. Instead, it offers two distinct fellowship programs that function as intensive, paid alternatives to conventional internships. Which one you target depends on whether you're drawn to frontier AI research or applied nonprofit work.

See all open positions on the Anthropic careers page. Note that the Fellows Program is research-focused and based in Berkeley/London, while Claude Corps is fully in-person at nonprofit host organizations across the US. Anthropic also occasionally posts Research Scientist Intern and Research Engineer Intern roles (PhD track), though these are rare and inconsistently available.

What Are the Eligibility Requirements?

Anthropic's eligibility requirements are deliberately broad, and that's by design. The company explicitly values execution ability over credentials:

• Education: No PhD required. No prior ML experience required. No published papers required. About half of Anthropic's technical staff had no prior ML experience, and "plenty of brilliant colleagues never went to college."

• Technical: Must be fluent in Python programming. This is non-negotiable. The CodeSignal assessment and all technical interview rounds are Python-focused.

• Work authorization: Must have existing authorization to work in the US, UK, or Canada. Anthropic does NOT sponsor visas for the Fellows Program.

• Availability: Full-time commitment (40 hours/week) for the entire 4-month duration. Must be physically located in the US, UK, or Canada during the program.

• Motivation: Genuine excitement about reducing catastrophic risks from advanced AI systems. This is assessed throughout the process and is a top-tier selection criterion.

Do I Actually Need Research Experience to Get In?

No.

And Anthropic is unusually direct about this. Their program page states: "We care much more about your ability to execute on research than your credentials." What does that mean in practice? It means a physics major who writes clean Python and can take an ambiguous problem and make concrete progress has a real shot, even without a single ML paper. The accepted backgrounds include computer science, mathematics, physics, cybersecurity, philosophy, economics, neuroscience, and "adjacent fields." Career changers are explicitly welcomed.

But don't confuse low credential requirements with low standards. Approximately 150 candidates complete all interview rounds for 32 spots per cohort. The bar is high; it's just measured differently than at a traditional research lab.

What Skills Does Anthropic Look For, and How Do You Build Them?

Seven real Anthropic program descriptions and job postings tell a clear story. Python appears in all seven. Machine learning fundamentals and AI safety alignment each show up in five.

But the soft-skill signal is equally strong: ambiguity tolerance and independent research execution score four out of seven, signaling that Anthropic hires people who can drive projects forward without hand-holding. What does that pattern reveal? They want builders who care about safety, not credential collectors who happen to know the right vocabulary.

How Is Demand for AI/ML Research Interns Moving Right Now?

Build These Skills Before You Apply

Every top skill in the chart maps to a remote Externship where you finish a real company project before the next cohort opens.

How tight is the overlap? The Wayfair deliverable is a working AI agent built with production Python and LLM integration, which maps directly to Anthropic's CodeSignal assessment (building systems under time pressure) and the LLM Engineering interview round (completing inference pipeline code). The Beats by Dre project produces quantitative analysis that shows the data reasoning and structured problem-solving the CodeSignal filter tests for.

What Is the Anthropic Application and Interview Process Like?

The Anthropic Fellows Program runs a five-stage funnel, and the rolling timeline means each stage starts within days of the previous one:

1. Submit your application on Greenhouse with your resume, optional code samples or publications, and three required references. Write substantive responses on motivation, research interests, and team fit. Don't rush these: Anthropic weighs your written framing of AI safety motivation heavily.

2. CodeSignal Online Assessment (90 minutes) arrives within days of advancing. One complex problem divided into 4 progressively harder levels (250 points each, 1,000 maximum). These aren't traditional algorithm puzzles; they're systems-building problems like implementing in-memory databases with TTL expiration. Official cutoff is 480, but practical cutoff is 600+.

3. Live Coding Screen (90 minutes) with an Anthropic engineer via CodeSignal. Logic-heavy implementation with dense specifications. Spend 10 to 15 minutes clarifying the problem before coding. Narrate your thought process throughout.

4. Final Loop: LLM Engineering (55 minutes) + Research Brainstorm (15 minutes). The engineering round uses Google Colab to complete missing LLM inference pipeline code. The brainstorm is an open-ended conversation about safety research framing with a potential mentor. This is where mission alignment gets tested directly.

5. References are contacted throughout the process (not just at the end). Your three references may be reached as early as the application review stage, so brief them in advance.

Important: Anthropic moves quickly between rounds. Expect scheduling within 3 days between stages. Total timeline from application to offer is approximately 3 to 6 weeks. No AI tools are permitted during the CodeSignal assessment.

And the entire process is designed to test practical engineering over theoretical knowledge, so practice building small systems quickly under time pressure rather than grinding isolated algorithm problems.

What Students on Reddit Say

Three community threads show what the process looks like from the inside.

How Do You Stand Out When Applications Are Reviewed on a Rolling Basis?

Three moves, all executable before the next cohort opens. First, apply the day the listing appears on Greenhouse. Rolling review plus limited cohort seats means timing is a filter in itself, so set alerts on the Anthropic careers page and check weekly. Second, build evidence in the fellowship's exact language: Python fluency, production-quality systems, and a proven project that shows you can take an ambiguous problem and make concrete progress. An Externship project building a real AI agent provides exactly this kind of artifact.

And third, study Anthropic's published research before you write your application essays. Reference Constitutional AI, mechanistic interpretability, or the responsible scaling policy with specific observations rather than generic praise. Anthropic screens for mission alignment at every stage, and candidates who can't articulate why AI safety matters to them personally don't advance regardless of technical strength.

What Other Companies Should You Consider?

Anthropic competes in the frontier AI lab tier. If you're targeting AI research fellowships and internships, these are the closest comparisons in mission, compensation, and selectivity.

Our tech internships summer 2027 guide maps the full landscape of AI lab and Big Tech research internship timelines in one place.

FAQ

Does Anthropic offer traditional internships?

No. As of 2026, Anthropic doesn't run a conventional internship program. Instead, they offer the Anthropic Fellows Program (a 4-month paid research fellowship) and the Claude Corps Fellowship (a 12-month nonprofit AI deployment). The Fellows Program is the closest equivalent to a research internship at other AI labs.

How much do Anthropic fellows get paid?

Anthropic Fellows receive $3,850 per week, which works out to roughly $15,400 per month. They also get approximately $15,000 per month in compute credits. Over the full 4 months, total compensation value is about $121,600. Claude Corps Fellows earn $85,000 per year with full benefits.

Do I need a PhD to apply to the Anthropic Fellows Program?

No. Anthropic explicitly states you don't need a PhD, prior ML experience, or published papers. They prioritize execution ability and genuine AI safety motivation over credentials. About half their technical staff had no prior ML experience before joining.

What programming language do I need to know?

Python is non-negotiable. You must be fluent in Python. The CodeSignal assessment and all technical interview rounds are Python-focused. Familiarity with PyTorch and ML frameworks is strongly valued but not required at application stage.

Can international students apply?

You must have existing work authorization in the US, UK, or Canada. Anthropic doesn't sponsor visas for the Fellows Program. If you need employment sponsorship of any kind, this fellowship is effectively unavailable unless you already hold valid authorization.

What is the acceptance rate for the Anthropic Fellows Program?

Approximately 150 candidates complete all interview rounds for around 32 spots per cohort, giving a final-round acceptance rate of about 21%. The overall application-to-offer rate is estimated at less than 5% because the CodeSignal assessment eliminates 60 to 70% of applicants.

Can the fellowship convert to a full-time job at Anthropic?

Yes. Over 40% of first-cohort fellows joined Anthropic full-time. The general conversion rate across cohorts runs 25 to 50%. Strong performance during the program signals good fit for permanent research roles, though conversion isn't guaranteed.

Is the fellowship remote or in-person?

The Anthropic Fellows Program offers shared workspaces in Berkeley, California and London, UK, with remote options for candidates in eligible countries. Claude Corps is fully in-person at nonprofit host organizations across the US, with a mandatory one-week orientation in San Francisco.

Applications are rolling and cohort spots are limited. Spend the runway building proof: a remote Externship turns "interested in AI safety" into a finished project that a rolling application can point at.