Anthropic Residency 2026: Up to $500K, No ML Background Needed

Anthropic will pay $350,000 to $500,000 a year (annualized) to bring professors, technical founders and staff-level engineers into AI safety research, and you do not need a machine learning background to apply. Placements run from six weeks to six months. Applications close on Sunday, November 22, 2026, which leaves you about two weeks. Here is who it fits, what you actually get, and how to build an application in time.

What the Anthropic Residency is

The Anthropic Residency is a paid, fixed term research program announced on Anthropic’s alignment blog in early November. It targets people who have already led serious technical work in another field and now want to move into full time AI safety and benefit research.

Anthropic’s pitch for why now is simple. Plenty of senior ML researchers, including many inside Anthropic, came from physics, math, neuroscience or software engineering. As models get more capable, choosing what is worth researching matters more, and senior people in quantitative fields tend to be good at exactly that.

The program builds on the Anthropic Fellows Program, adapted for people further along in their careers. Same idea, bigger pay, shorter runway options.

What you get

Pay: $350,000 to $500,000 a year equivalent, according to Anthropic.
Length: six weeks, three months or six months, with flexible start dates.
A real project: you work on problems outlined by Anthropic researchers, matched to your background and interests.
A mentor: one Anthropic researcher paired with you, with regular one on one time.
Ramp up time: dedicated self study time, talks and guidance on modern ML and AI safety. Anthropic says you are not expected to arrive as an ML expert.
Compute: access to frontier models and a dedicated compute budget.
Location: San Francisco preferred. Remote is possible, especially for shorter placements.

The stated goal: by the end, you can keep doing AI research full time, at Anthropic or somewhere else.

What does the pay work out to? The page says annualized equivalent, and one outlet describes it as prorated. If that is how it works, here is my arithmetic (not an Anthropic table):

Placement	Rough total at $350K a year	Rough total at $500K a year
Six weeks	About $40,000	About $58,000
Three months	About $88,000	About $125,000
Six months	About $175,000	About $250,000

Who it is built for

Anthropic names three example profiles: a professor in a quantitative discipline, a technical founder, or a staff-level engineer. The candidate list is more specific. You should be someone who:

Has led ambitious work in a technical field
Cares about AI being safe and beneficial for society
Wants to move into full time empirical AI research
Does well in fast, collaborative settings
Can implement ideas quickly and explain them clearly

No PhD, no prior ML experience and no published papers are required. Anthropic also says it particularly encourages applications from groups underrepresented in tech.

What you would actually research

Anthropic groups the work into three buckets.

AI safety and alignment: scalable oversight, adversarial robustness and AI control, model organisms of misalignment, interpretability, safeguards research and AI welfare.
AI security and frontier red teaming: stress testing frontier models in areas like cybersecurity, robotics and cryptography.
ML systems and reinforcement learning: improving frontier model training and its infrastructure, from architectures and optimizers to training data, accelerator simulators and RL environment pipelines.

To get a feel for the output, look at what past Fellows produced: subliminal learning, the Petri auditing tool and persona vectors.

Which mentor matches your background

Anthropic lists nine potential mentors, and the backgrounds are the interesting part. These are potential mentors, not guaranteed matches.

If you come from	Look at	Their focus
Physics or neuroscience	Jascha Sohl-Dickstein	Safety and pretraining
Theoretical neuroscience	Jack Lindsey	Model Psychology team lead
Mathematics	Samuel Marks	Scalable Oversight team lead
Biomedical informatics	Elana Simon	Interpretability of language and biological models
Security or adversarial ML	Nicholas Carlini	Adversarial ML and AI security
NLP	Belinda Li	How models represent the world, users and themselves
Data science	Sara Price	Alignment Training team lead
Machine learning	Ethan Perez	Alignment Science team lead
Independent AI safety research (such as MIRI)	Evan Hubinger	Alignment Stress-Testing team lead

Residency vs Fellows: which one is yours

If you are earlier in your career, the Fellows Program is the better door. Anthropic says Residency applicants may also be considered for it. I compared a similar program in my post on the Perplexity Research Fellowship.

	Residency	Fellows Program
Aimed at	Established leaders: professors, founders, staff engineers	Promising technical talent, regardless of previous experience
Length	Six weeks, three months or six months	Four months
Pay	$350K to $500K a year equivalent	$3,850 a week (about $200K annualized, my math)
Compute	Dedicated budget, amount not stated	About $15K a month
Location	San Francisco preferred, remote possible	Berkeley or London, or remote in the US, UK and Canada

Put plainly, the Residency pays roughly 1.75 to 2.5 times the Fellows weekly rate, annualized. That gap is the price of asking senior people to step away from senior jobs.

Your 16 day application plan

Today is November 6. The deadline is November 22, and the listing on Index Ventures’ job board gives it as 11:59 PM Pacific. Anthropic does not publish a required application format on the announcement page, so treat the steps below as my suggested plan, not official requirements.

Days 1 to 2: pick your lane. Choose one of the three research buckets and one potential mentor from the table above. A narrow target beats a vague “I want to work on safety.”
Days 3 to 5: read three things. Start with Anthropic’s recommended research directions, then one paper from your lane, then one past Fellows project. Write down one question you would want to test.
Days 6 to 9: build a small proof. Run a tiny experiment on a public or open model, then write one page on what you tried, what you found and what broke. Anthropic wants people who implement fast and communicate clearly. A short write up shows both.
Days 10 to 12: write your bridge story. Your hardest technical win, why the skill transfers, and why you want to work on safety now. Keep it concrete. Numbers and outcomes beat adjectives.
Days 13 to 14: sort the logistics. Pick a duration, check what leave or sabbatical you can get, and read the full US posting or UK posting for terms the announcement does not cover.
Days 15 to 16: submit early. Do not leave it for the last night. Portals fail, and a Sunday deadline is easy to misjudge across time zones.

Honest limitations

This is a strong offer. It is also a recruiting page, so read it with a few caveats in mind.

Everything here is Anthropic’s own account. Coverage elsewhere largely repeats the announcement. I could not open the full job posting, so I cannot confirm benefits, visa terms or the interview process. Read the posting before you plan around any of it.
The announcement page shows no author or date. Job listings and news coverage point to applications opening around November 2 to 3.
No numbers on competition. The page does not say how many Residents Anthropic will take or how many people are expected to apply.
The pay wording is loose. “Annualized equivalent” does not tell you whether a six week placement pays a prorated slice or something else, or what benefits come with it.
Six weeks is a tryout, not a career. The stated goal is that you can continue AI research somewhere. It is not a promised job offer. Leaving a tenured post or a company for a short placement has real costs.
The projects are Anthropic’s. You work on problems outlined by their researchers. That suits some people and frustrates others.

Quick answers

Do I need a PhD or ML experience?
No. Anthropic says no PhD, no prior ML experience and no published papers are required.

How much does the Anthropic Residency pay?
$350,000 to $500,000 a year equivalent, according to the announcement.

How long is it?
Six weeks, three months or six months.

Can I do it remotely?
San Francisco is preferred. Remote work is possible, especially for shorter placements.

When is the deadline?
November 22, 2026.

Sources

Know a professor, founder or senior engineer who has been circling AI safety? Send them this before November 22.