Exploring the Anthropic Fellows Program: AI Research Opportunities in 2026

Exploring the Anthropic Fellows Program: A Deep Dive into AI Research Opportunities

The field of Artificial Intelligence is rapidly advancing, and with it comes a growing need for skilled researchers and engineers dedicated to building safe and beneficial AI systems. Anthropic, a company focused on creating reliable, interpretable, and steerable AI, offers a unique opportunity for aspiring AI professionals through its Anthropic Fellows Program. This program provides funding, mentorship, and hands-on research experience, aiming to cultivate talent regardless of prior experience. The program is designed to foster empirical research projects aligned with Anthropic’s priorities, with a strong emphasis on producing public outputs like research papers.

The Anthropic Fellows Program is structured to offer a comprehensive and immersive experience in AI research. It spans four months of full-time work, providing fellows with direct mentorship from experienced Anthropic researchers. Participants gain access to shared workspaces in Berkeley, California, or London, UK, and become connected to the wider AI safety and security research community. The program also offers a weekly stipend, compute funding, and support for other research expenses, making it an accessible pathway into cutting-edge AI research.

Program Overview and Goals

Anthropic’s mission is to ensure that AI is developed responsibly and benefits society. The Fellows Program is a key initiative in achieving this mission by nurturing new talent in AI research and engineering. The program encourages fellows to use external infrastructure, such as open-source models and public APIs, to conduct empirical projects. A significant goal is for fellows to produce a public output, such as a paper submission, with a history of over 80% of fellows achieving this in previous cohorts. The program accepts applications on a rolling basis for cohorts starting in July 2026 and beyond, with flexibility for those who cannot meet the standard September start date.

What to Expect as a Fellow

Participants in the Anthropic Fellows Program can anticipate a structured yet flexible research environment. The program typically lasts for four months of full-time engagement, though extensions may be possible. Fellows receive direct guidance from Anthropic researchers, who act as mentors throughout the project lifecycle. The program provides access to physical workspaces in Berkeley or London, fostering collaboration and interaction with peers and mentors. Beyond the workspace, fellows are integrated into the broader AI safety and security research community.

Financially, the program offers a competitive weekly stipend of $3,850 USD, £2,310 GBP, or $4,300 CAD, along with country-specific benefits. Significant funding is also allocated for compute resources, estimated at around $15,000 USD per month, and other necessary research expenses. This financial support allows fellows to focus on their research without undue financial pressure.

The Application and Interview Process

The application process for the Anthropic Fellows Program is designed to identify promising candidates who are passionate about AI safety and possess strong technical foundations. It begins with an initial application review, followed by reference checks. Candidates who pass this stage will proceed to technical assessments and interviews. A final research discussion will help determine the best fit for the program and specific workstreams.

Anthropic strongly encourages applications from individuals who may not meet every single qualification listed. They recognize that underrepresented groups often experience imposter syndrome and may doubt their suitability. The company emphasizes that diverse perspectives are crucial for developing AI systems with significant social and ethical implications. Therefore, anyone interested in the work is urged to apply, as Anthropic seeks to build a team that reflects a wide range of backgrounds and viewpoints.

Compensation and Benefits

The Anthropic Fellows Program offers a substantial weekly stipend, set at $3,850 USD, £2,310 GBP, or $4,300 CAD. This compensation is for a commitment of approximately 40 hours per week over the four-month program duration, with the possibility of extension. While the program does not guarantee full-time employment offers, strong performance can lead to consideration for future roles at Anthropic. Historically, between 25% and 50% of fellows have received full-time offers, and many others have gone on to contribute to AI safety and security work at other organizations.

Diverse Workstreams for Specialized Research

The Anthropic Fellows Program has expanded to include various workstreams, catering to different areas of AI research and engineering. While there is significant overlap in skills and responsibilities across these streams, candidates can indicate their preferences. The primary workstreams include:

AI Safety Fellows

This stream focuses on ensuring AI systems are safe and beneficial for society. Research areas include Scalable Oversight, developing techniques for managing highly capable AI; Adversarial Robustness and AI Control, creating methods to keep AI systems safe in unfamiliar situations; Model Organisms, studying misalignment to understand failure modes; Model Internals/Mechanistic Interpretability, deciphering how large language models work; and AI Welfare, improving understanding and evaluation of AI well-being. Potential mentors include prominent researchers like Sam Bowman and Jascha Sohl-Dickstein.

AI Security Fellows

This workstream is dedicated to reducing catastrophic risks from advanced AI systems through security research. Projects may involve contributions to open-source LLM or security repositories, tackling ambiguous technical problems, and engaging in offensive security work like pentesting. Mentors such as Nicholas Carlini and Keri Warr guide fellows in areas like identifying smart contract exploits and strengthening red team capabilities.

ML Systems & Performance Fellows

This stream emphasizes the engineering aspects of AI, focusing on building and optimizing ML systems. Projects can involve developing CPU simulators, adding support for different accelerators, creating on-demand infrastructure, and building complex data pipelines. Strong software engineering skills and experience with large-scale distributed systems are highly valued. Mentors like Alwin Peng and Zygi Straznickas lead projects in this area.

Reinforcement Learning Fellows

Fellows in this stream work on advancing reinforcement learning (RL) techniques for AI development. Projects may include building tools to analyze training data, researching generalization in AI, creating RL environments for improving AI models, and implementing RL algorithms for safety-related tasks. Expertise in software engineering and ML systems is beneficial.

Economics Fellows

This workstream, part of The Anthropic Institute, focuses on the economic and societal impacts of AI. Projects involve designing and conducting empirical research on AI’s economic effects, developing new methods for studying AI’s impact on labor markets, and analyzing the offense-defense balance for AI-enabled capabilities. Mentors include economists and policy experts like Maxim Massenkoff and Jack Clark.

Logistics and Requirements

To participate in the Anthropic Fellows Program, candidates must possess work authorization in the US, UK, or Canada and be located in one of these countries during the program. While shared workspaces are available in London and Berkeley, the program also supports remote fellows across the US, UK, and Canada. It is important to note that Anthropic does not currently sponsor visas for fellows. The program duration is four months full-time, but flexibility is offered for those unable to commit to the entire period. Applications and interviews are managed by Constellation, Anthropic’s recruiting partner.

How Anthropic Stands Out

Anthropic distinguishes itself by viewing AI research as “big science,” focusing on large-scale, cohesive research efforts rather than smaller, isolated problems. The company prioritizes impact, aiming to advance steerable and trustworthy AI. Anthropic treats AI research as an empirical science, drawing parallels with physics and biology. Collaboration is central to their approach, with frequent research discussions ensuring the pursuit of high-impact work. Strong communication skills are highly valued, reflecting the collaborative nature of their research. Their research directions often build upon previous work in areas like GPT-3, interpretability, and scaling laws.

Frequently Asked Questions