Anthropic Residency 開放申請，招募資深技術領袖

Today we are opening applications for the Anthropic Residency, our paid, fixed-term research program with placements lasting between six weeks and six months.

We are looking for people who have led ambitious work in a variety of technical fields and want to transition to full-time work focused on making AI systems safe and beneficial for all of humanity. If you are a professor in a quantitative discipline, a technical founder, or a staff-level engineer and want to transition to work on transformative AI, we want to hear from you. Background in machine learning is not required.

Applications for this round close on November 22, 2026. Apply here (US) or here (UK).

Why we’re opening this round

Machine learning is a young field. Many of today’s senior researchers, including many at Anthropic, started out as physicists, mathematicians, neuroscientists, or software engineers. As model capabilities improve, what matters most in AI research is increasingly the ability to frame a research agenda and judge what is worth doing. Senior people in quantitative fields often already excel at these skills; the Residency helps them build the understanding of frontier AI they need to apply those skills in this new field.

This round builds on the Anthropic Fellows Program, which demonstrated how quickly new researchers can contribute to the field. Past fellows have produced work on subliminal learning, the Petri auditing tool, and persona vectors. This round adapts that model for established technical leaders.

What to expect in this round

Work on critical research problems: Residents will work on problems outlined by Anthropic researchers, and matched to the Resident’s interests and background.
Mentorship from leading researchers: Each Resident will be paired with an Anthropic mentor and receive regular one-on-one guidance.
Support in transitioning to the field: We don’t expect Residents to arrive as ML experts. The Residency includes dedicated self-study time, talks, and guidance to help them get up to speed on modern ML and AI safety research.
Access to compute: Residents will work with frontier models and have a dedicated compute budget for their projects.
Flexible timing: Placements run for six weeks, three months, or six months, and we will ask about your preferred duration when you apply. We will aim to arrange flexible start dates.
Location: We prefer that Residents work from our San Francisco offices, but remote participation is possible, especially for shorter placements. Most full-time research roles at Anthropic require relocation to San Francisco.
Compensation that reflects seniority: Residents will be paid $350,000 to $500,000/year equivalent.

By the end of the Residency, we aim for every Resident to be in a position to continue working on AI research full-time, whether at Anthropic or elsewhere.

Mentors and research areas

Potential Anthropic mentors include:

Sara Price: leads the Alignment Training team; background in data science
Ethan Perez: leads the Alignment Science team; background in machine learning
Evan Hubinger: leads the Alignment Stress-Testing team; formerly at The Machine Intelligence Research Institute (MIRI)
Samuel Marks: leads the Scalable Oversight team; background in mathematics
Elana Simon: researches the interpretability of language models and biological models; background in biomedical informatics
Belinda Li: researches how language models represent the world, users, and themselves; background in natural language processing (PhD, MIT)
Jack Lindsey: leads the Model Psychology team; background in theoretical neuroscience
Nicholas Carlini: researches adversarial machine learning and AI security; previously at DeepMind and Google Brain
Jascha Sohl-Dickstein: works at the intersection of safety and pretraining; background in physics and neuroscience, formerly Principal Scientist at Google Brain

Residents will work on empirical research questions in collaboration with Anthropic’s research teams. Areas we are particularly interested in include:

AI safety & alignment

Scalable Oversight: Developing techniques to keep highly capable models helpful and honest, even as they surpass human-level intelligence in various domains.^[1]
Adversarial Robustness and AI Control: Creating methods to ensure advanced AI systems remain safe and harmless in unfamiliar or adversarial scenarios.^[2]
Model Organisms: Creating model organisms of misalignment to improve our empirical understanding of how alignment failures might arise.^[3]
Model Internals and Interpretability: Advancing our understanding of the internal workings of large language models to enable more targeted interventions and safety measures.^[4]
Safeguards Research: Developing robust defenses against adversarial attacks, comprehensive evaluation frameworks for model safety, and automated systems to detect and mitigate potential risks before deployment.^[5]
AI Welfare: Improving our understanding of potential AI welfare and developing related evaluations and mitigations.^[6]

AI security & frontier red-teaming

Stress-testing frontier models to understand the full extent of their current capabilities in fields such as cybersecurity, robotics, or cryptography.^[7]

ML systems & reinforcement learning

Improving how frontier models are trained and building the infrastructure behind that research, from model architecture, optimizers, and training data to accelerator simulators, hardware backends, and RL environment pipelines.

For a fuller picture of the questions we care about, see our Recommendations for Technical AI Safety Research Directions and recent publications from our Alignment, Interpretability, and Frontier Red teams.

We are looking for people who

Have led ambitious work in a technical field (a professor in a quantitative discipline, a technical founder, or a staff-level engineer)
Are motivated by making sure AI is safe and beneficial for society as a whole
Are excited to transition into working on empirical AI research full-time
Thrive in fast-paced, collaborative environments
Can implement ideas quickly and communicate clearly

You don’t need a PhD, prior ML experience, or published papers.

How to apply

To be considered for this round, please apply via the US or UK posting by November 22, 2026.

We value candidates who bring unique perspectives and experiences to AI research. We particularly encourage applications from underrepresented groups in tech. Applicants may also be considered for the Anthropic Fellows Program.

Links to past research

Representative research from the Anthropic Fellows Program:

[1] Automated Weak-to-Strong Researcher, Removing Sandbagging in LLMs by Training with Weak Supervision

[2] SHADE-Arena: Evaluating Sabotage and Monitoring in LLM Agents, SLEIGHT-Bench: Finding Blind Spots in AI Monitors

[3] AuditBench: Evaluating Alignment Auditing Techniques on Models with Hidden Behaviors, Subliminal Learning: Language Models Transmit Behavioral Traits via Hidden Signals in Data

[4] Natural Language Autoencoders Produce Unsupervised Explanations of LLM Activations, Mechanisms of Introspective Awareness

[5] Poisoning Fine-tuning Datasets of Constitutional Classifiers, Detecting Adversarial Fine-tuning with Auditing Agents

[6] The LLM Has Left The Chat: Evidence of Bail Preferences in Large Language Models, Gemma Needs Help: Investigating and Mitigating Emotional Instability in LLMs

[7] AI Agents Find $4.6M in Blockchain Smart Contract Exploits, How Claude Performs on Robotics Tasks