Anthropic AI Safety Fellowship 2026: How to apply, $15,000 funding, duration & hiring chances
Anthropic, founded by former OpenAI researchers, has positioned itself as one of the leading firms focused on AI alignment and “steerable” AI systems. The company says its mission is to create AI that remains reliable, interpretable, and beneficial for society.
- May 14, 2026,
- Updated May 14, 2026 10:02 PM IST
As competition in artificial intelligence intensifies globally, AI startup Anthropic is significantly expanding its highly watched Fellows Program, offering aspiring researchers and engineers a pathway into frontier AI safety, security, reinforcement learning, and policy research.
The San Francisco-headquartered company has opened rolling applications for new cohorts expected to begin in late September, with the program positioned as a talent accelerator for individuals seeking hands-on experience in advanced AI systems and alignment research.
MUST READ | Anthropic’s AI takeover? Indian-origin CFO Krishna Rao says Claude now handles 90% of coding work
Anthropic said the initiative is designed to “foster AI research and engineering talent” by funding and mentoring promising technical candidates, even if they do not come from traditional AI industry backgrounds.
4-month fellowship with research funding
The four-month full-time fellowship offers direct mentorship from Anthropic researchers, access to shared workspaces in Berkeley and London, compute funding of roughly $15,000 per month, and weekly stipends of $3,850 in the United States, £2,310 in the United Kingdom, and CAD 4,300 in Canada.
The company noted that more than 80% of fellows in earlier cohorts produced research papers or public technical outputs, highlighting the program’s growing influence within the AI research ecosystem.
DON'T MISS | OpenAI takes on Anthropic’s Claude Mythos with new cybersecurity AI system ‘Daybreak’
Anthropic said fellows will primarily work using external infrastructure, including open-source AI models and public APIs, with the goal of creating publishable research or practical technical outputs.
AI safety becomes a strategic priority
The expansion reflects a broader shift in the global AI industry, where companies are racing not only to build more capable large language models, but also to strengthen safeguards around increasingly powerful systems.
Anthropic, founded by former OpenAI researchers, has positioned itself as one of the leading firms focused on AI alignment and “steerable” AI systems. The company says its mission is to create AI that remains reliable, interpretable, and beneficial for society.
DO CHECKOUT | Pentagon deploys Anthropic’s Mythos AI cyber model despite plans to phase out company’s tools
The new fellowship structure spans multiple specialised tracks, including AI Safety, AI Security, ML Systems & Performance, Reinforcement Learning, and Economics & Policy.
Focus areas include alignment, cybersecurity & RL
Under the AI Safety Fellows stream, participants will work on research areas such as scalable oversight, adversarial robustness, mechanistic interpretability, and AI welfare.
Anthropic said the work aims to better understand how advanced AI systems behave internally and how catastrophic risks from future AI models could be mitigated.
Potential mentors in the safety stream include researchers such as Sam Bowman, Alex Tamkin, Ethan Perez, and Jascha Sohl-Dickstein.
The AI Security Fellows track focuses on offensive security research, red teaming, vulnerability discovery, and AI-enabled cyber risks. Anthropic highlighted earlier projects where AI agents reportedly identified millions of dollars worth of blockchain smart contract vulnerabilities.
Meanwhile, the Reinforcement Learning and ML Systems tracks are aimed at candidates with strong engineering and distributed systems expertise, particularly those interested in training, evaluating, and debugging large language models at scale.
Economics & Policy research added
Anthropic is also broadening the initiative into economics and public policy research through what it calls the Anthropic Institute Fellows program.
The company said projects may examine AI’s effects on labour markets, economic systems, cybersecurity, and societal resilience, alongside research into AI governance and early warning signals for future AI risks.
Unlike many elite AI research programs that focus heavily on academic credentials, Anthropic stressed that the fellowship is open to applicants from diverse backgrounds and encouraged candidates from underrepresented groups to apply even if they do not meet every listed qualification.
No guarantee for jobs
Anthropic clarified that the fellowship does not guarantee a full-time role at the company, though previous cohorts reportedly saw between 25% and 50% of fellows receive employment offers.
The company also acknowledged the rapid rise of recruitment scams in the AI sector, warning applicants that official communications would only come from verified Anthropic-linked channels or approved recruiting partners.
As competition in artificial intelligence intensifies globally, AI startup Anthropic is significantly expanding its highly watched Fellows Program, offering aspiring researchers and engineers a pathway into frontier AI safety, security, reinforcement learning, and policy research.
The San Francisco-headquartered company has opened rolling applications for new cohorts expected to begin in late September, with the program positioned as a talent accelerator for individuals seeking hands-on experience in advanced AI systems and alignment research.
MUST READ | Anthropic’s AI takeover? Indian-origin CFO Krishna Rao says Claude now handles 90% of coding work
Anthropic said the initiative is designed to “foster AI research and engineering talent” by funding and mentoring promising technical candidates, even if they do not come from traditional AI industry backgrounds.
4-month fellowship with research funding
The four-month full-time fellowship offers direct mentorship from Anthropic researchers, access to shared workspaces in Berkeley and London, compute funding of roughly $15,000 per month, and weekly stipends of $3,850 in the United States, £2,310 in the United Kingdom, and CAD 4,300 in Canada.
The company noted that more than 80% of fellows in earlier cohorts produced research papers or public technical outputs, highlighting the program’s growing influence within the AI research ecosystem.
DON'T MISS | OpenAI takes on Anthropic’s Claude Mythos with new cybersecurity AI system ‘Daybreak’
Anthropic said fellows will primarily work using external infrastructure, including open-source AI models and public APIs, with the goal of creating publishable research or practical technical outputs.
AI safety becomes a strategic priority
The expansion reflects a broader shift in the global AI industry, where companies are racing not only to build more capable large language models, but also to strengthen safeguards around increasingly powerful systems.
Anthropic, founded by former OpenAI researchers, has positioned itself as one of the leading firms focused on AI alignment and “steerable” AI systems. The company says its mission is to create AI that remains reliable, interpretable, and beneficial for society.
DO CHECKOUT | Pentagon deploys Anthropic’s Mythos AI cyber model despite plans to phase out company’s tools
The new fellowship structure spans multiple specialised tracks, including AI Safety, AI Security, ML Systems & Performance, Reinforcement Learning, and Economics & Policy.
Focus areas include alignment, cybersecurity & RL
Under the AI Safety Fellows stream, participants will work on research areas such as scalable oversight, adversarial robustness, mechanistic interpretability, and AI welfare.
Anthropic said the work aims to better understand how advanced AI systems behave internally and how catastrophic risks from future AI models could be mitigated.
Potential mentors in the safety stream include researchers such as Sam Bowman, Alex Tamkin, Ethan Perez, and Jascha Sohl-Dickstein.
The AI Security Fellows track focuses on offensive security research, red teaming, vulnerability discovery, and AI-enabled cyber risks. Anthropic highlighted earlier projects where AI agents reportedly identified millions of dollars worth of blockchain smart contract vulnerabilities.
Meanwhile, the Reinforcement Learning and ML Systems tracks are aimed at candidates with strong engineering and distributed systems expertise, particularly those interested in training, evaluating, and debugging large language models at scale.
Economics & Policy research added
Anthropic is also broadening the initiative into economics and public policy research through what it calls the Anthropic Institute Fellows program.
The company said projects may examine AI’s effects on labour markets, economic systems, cybersecurity, and societal resilience, alongside research into AI governance and early warning signals for future AI risks.
Unlike many elite AI research programs that focus heavily on academic credentials, Anthropic stressed that the fellowship is open to applicants from diverse backgrounds and encouraged candidates from underrepresented groups to apply even if they do not meet every listed qualification.
No guarantee for jobs
Anthropic clarified that the fellowship does not guarantee a full-time role at the company, though previous cohorts reportedly saw between 25% and 50% of fellows receive employment offers.
The company also acknowledged the rapid rise of recruitment scams in the AI sector, warning applicants that official communications would only come from verified Anthropic-linked channels or approved recruiting partners.