Anthropic Opens Residency Researcher Program, Targeting Senior Technical Talent Transitioning into AI Safety
Anthropic opened applications today for Anthropic Residency (Residency Research Program). This is a paid, fixed-term full-time empirical research program lasting six weeks, three months, or six months, aimed at people who have already achieved success in other technical fields and hope to transition into full-time AI safety research. A machine learning background is not required.
The company explicitly names three types of applicants who are a fit: professors in quantitative disciplines such as mathematics, physics, and statistics; technical founders; and Staff-level engineers. Applicants do not need a doctoral degree, machine learning experience, or a publication record.
Each participant is assigned an Anthropic researcher as a mentor, with regular one-on-one guidance provided. Topics are proposed by Anthropic researchers and matched according to individual interests and backgrounds. The program includes self-study time, lectures, and tutoring to help participants make up for gaps in modern machine learning and AI safety fundamentals; participants can use frontier models and have a dedicated compute budget.
The research directions mainly include: AI safety and alignment (such as scalable oversight, interpretability, and AI welfare); offensive and defensive AI safety work and frontier red team testing; and model training and the underlying infrastructure (architecture, optimizers, training data, accelerator simulators, hardware backends, and reinforcement learning environments).
For U.S. positions, the annual salary is the equivalent of US$350,000 to US$500,000, prorated according to the actual length of the program. Anthropic prefers participants to work in the San Francisco office, but remote work is supported, especially for shorter programs. Visas can be arranged for three months and above. The application deadline is November 22, 2026 (11:59 p.m. Pacific Time). The goal is for every participant to be able to continue doing AI research full-time after the program ends.
Original source:
Title: Anthropic Residency: applications open for established technical leaders Publisher: Anthropic Alignment Science Blog Publication date: 2026-11-02 URL: https://alignment.anthropic.com/2026/anthropic-residency/ Source nature: official company announcement (Alignment Science Blog) Verification status: full official announcement read for this article (verified on 2026-11-02) Supporting content: supports all core facts in the main text, including the program’s nature and duration, no ML background requirement, target groups (professors in quantitative disciplines, technical founders, Staff-level engineers), mentorship and research directions, U.S. salary equivalent of US$350,000 to US$500,000, remote work and visa arrangements, and the November 22 deadline
ABAB AI Insight
This program is an adaptation based on Anthropic’s Fellows program. Fellows is aimed at researchers new to the field and has produced results such as subliminal learning, the Petri model auditing tool, and persona vectors. Residency applies a similar model to established technical leaders, emphasizing that the ability to determine research directions and judge what is worth doing matters more than pure ML experience. Anthropic notes that many current senior researchers originally came from physics, mathematics, neuroscience, or software engineering.
Potential mentors include Ethan Perez, head of the Alignment Science team; Jack Lindsey, head of the Model Psychology team; Jascha Sohl-Dickstein, former chief scientist at Google Brain; and Nicholas Carlini, who researches adversarial machine learning and AI safety.
The compensation, converted on an annualized basis to US$350,000 to US$500,000, reflects the positioning of senior talent; shorter periods support remote work, and visas can be sponsored for programs of three months or longer. Anthropic says that most full-time research roles still require relocation to San Francisco, while this program offers a transition path. The application process is expected to include an initial screening, technical evaluation, research discussion, and background check, with offer notifications sent in January 2027; applicants may also be considered for the Fellows program.
This move reflects how frontier AI labs, amid a shortage of safety talent, are proactively bringing in experienced leaders from adjacent technical fields rather than relying only on traditional ML paths to develop talent.
ABAB News · Cognitive Law
No machine learning background, doctoral degree, or papers are required; what matters is the existing qualifications of professors in quantitative disciplines, technical founders, or Staff-level engineers.
The annualized salary for U.S. positions is equivalent to US$350,000 to US$500,000; the shorter the duration, the less actual take-home pay.
The program is adapted from the Fellows model, emphasizes the ability to judge research directions, and aims for participants to be able to do AI research full-time after it ends.