Currently 22 active ai and ai safety grants, fellowships and residencies open to applicants in the US. Hand-curated and updated weekly. Almost every entry is funded; a few notable unpaid open calls and festival submissions are included as clearly flagged exceptions. Browse the list below, or use the interactive desk for filtering and shortlisting.
Subscribe: RSS feed · Calendar (.ics)
Anthropic's four-month STEM Fellows Program brings mathematicians, physicists, biologists, chemists, engineers and other domain scientists in to work on unsolved problems in their own fields using Claude. Fellows join small pods of peers from related disciplines, define hard open problems that Claude currently struggles with, build datasets, simulations and evaluations, test how the model plans experiments and interprets data, and map where its answers are confidently wrong, with access to frontier models and internal tools and weekly meetings with Anthropic researchers. Some fellowships may be extended. ELIGIBILITY: a STEM PhD plus several years of experience (for example postdoc or faculty), hands-on experience with Claude and Claude Code, full-time availability, and existing US work authorisation (no visa sponsorship); prior machine-learning experience helps but is not required. Runs 12 January to 8 May 2027, full time, San Francisco by default with a required one-week in-person orientation there; remote within the US may be considered case by case. Later stages include a take-home exercise and conversations with the research team. DEADLINE: 28 October 2026 (no time zone stated); offers planned by 20 November 2026.
The Foundation for American Innovation's Conservative AI Policy Fellowship is an eight-week, part-time, work-compatible programme on AI policy for conservative policy professionals in Washington DC. WHO: congressional and executive branch staff, policy researchers and think tank staff, tech professionals moving into AI policy, and exceptional early-career applicants; no technical AI background is required. FORMAT: weekly in-person sessions at FAI's Capitol Hill office (usually Fridays 12:00-13:30 or Tuesdays 18:30-20:30), about 2-3 hours a week, plus a required weekend retreat in February. WHAT YOU GET: a USD 1,500 stipend for eligible fellows (subject to ethics rules for government employees), with meals and retreat costs covered. APPLICATION: online form linked from the FAI fellowships page. DEADLINE: 30 October 2026.
Umbrella request for proposals for ambitious AI-first projects that make the transition to capable AI go well and use it to unlock science. Three connected layers, each with its own detailed call: I. Local Compute (compute that individuals or communities physically own and control); II. Coordination and Accountability (supercollaboration and decentralized alignment, human empowerment, AI insurance and open governance); III. AI-First Science: Bio, Neuro, Nano (automated nanotechnology, whole-brain emulation, replacement, enhancement, biostasis). AI-first means AI is the primary engine of progress, not tapped onto an existing research program; strong proposals identify narrow bottlenecks and propose something that can be built, tested or measured. Selected projects receive grant funding, private local compute, workspace at the San Francisco or Berlin AI Nodes, travel-paid field-building events and access to Foresight's advisors and scientific network. ELIGIBILITY: individuals, teams, and nonprofit or for-profit organisations worldwide (for-profits must justify their need for grant funding); applicants who want to be active in-person contributors at the Nodes are strongly prioritised. Grants typically $30,000 to $100,000, smaller for the Human Empowerment and AI Insurance & Open Governance areas and higher for the rest. Review takes about three months after the deadline. Deadline 31 October 2026, 23:59 PDT.
The MATS Residency, announced in August 2026 as a new track alongside the seasonal MATS fellowship, funds 6 to 24 months of high-independence AI safety research for experienced researchers, whether already in AI safety or moving in from adjacent fields. COMPENSATION: USD 155,000 to 285,000 per year gross depending on location, seniority and experience, plus insurance, retirement plan, PTO (17 days US / 20 days UK), conference funding, catered meals, and up to USD 25,000 in relocation support. ELIGIBILITY: at least one substantial piece of research-caliber work; credentials are not required. Full-time (part-time case by case for academics on sabbatical), in person at MATS offices in Berkeley, London, or Washington DC, with visa support where required. TIMELINE: the Winter 2027 intake window runs 25 August to 31 October 2026 (AoE), with decisions by end of November and interviews plus a paid work test in the first week of December; three intake windows per year.
Track I of Foresight Institute's AI for Science & Safety Nodes RFP funds building and scaling local compute setups that individuals or communities physically own and control. With five companies holding 71 percent of the world's AI compute, researchers relying on APIs and rented compute face data exposure, steep price increases and the risk of losing access altogether; open models are ready for serious science and safety work, but the ecosystem lacks sovereign and secure compute infrastructure. LOOKING FOR: local compute clusters in different sizes and configurations that are easy to set up and run across varied connectivity and electricity environments; blueprints for easy yet secure cluster management; models for bringing down cost, like selling idle compute or connecting to community clusters; ideally a business case for building, selling and distributing such clusters at scale. EXAMPLE DELIVERABLES: a prototype AI cluster an average person can deploy at home, software that runs a node without expert knowledge, economic models showing how the hardware pays for itself, a pilot deployment for individuals or communities. Selected projects receive grant funding, private local compute, workspace at the San Francisco or Berlin AI Nodes, travel-paid field-building events and access to Foresight's advisors and network; in-person contributors at the Nodes are strongly prioritised. Individuals, teams, and nonprofit or for-profit organisations worldwide may apply (for-profits must justify their need for grant funding). Deadline 31 October 2026, 23:59 PDT.
Track II of Foresight Institute's AI for Science & Safety Nodes RFP funds ways for increasingly capable AI systems to coordinate with each other, the people using them, and the wider world, without handing power to a single model, lab or government. THREE FOCUS AREAS: (1) Supercollaboration and decentralized alignment: coordination and credit-assignment protocols, tooling for massive human-AI collectives, incentive designs robust to gaming, mutual monitoring and cross-checking among agents, scalable verification, and open coordination protocols built on boundaries as Schelling points for multi-agent coordination. (2) Human Empowerment: evidence, decision rules and working tools for when human control genuinely empowers people versus when AI delegation is preferable, including a 'when to defer control' framework and mechanisms to detect and fix hollow oversight. (3) AI Insurance and Open Governance: building a middle layer between lab self-governance and government regulation, such as ways for insurers to judge and price frontier-lab risk, independent incident reporting and assessment that does not leak lab secrets, and mechanisms like developer mutuals, catastrophe bonds or liability for near-misses. Selected projects receive grant funding, private local compute, workspace at the San Francisco or Berlin AI Nodes, travel-paid field-building events and access to Foresight's advisors and network; in-person contributors at the Nodes are strongly prioritised. Individuals, teams, and nonprofit or for-profit organisations worldwide may apply (for-profits must justify their need for grant funding). Deadline 31 October 2026, 23:59 PDT.
Track III of Foresight Institute's AI for Science & Safety Nodes RFP funds AI-first science in bio, neuro and nano: work that asks what important science is stuck, what it is actually stuck on, and whether we can build the thing that removes that constraint. THREE FOCUS AREAS: (1) Automated nanotechnology: AI methods for nanoscale design, simulation and assembly, closed-loop fabrication and characterization tools, credible application pathways from toy problems to medicine, environment and bottom-up production, and defensive applications and safety frameworks. (2) Whole-brain emulation: cheaper and faster brain mapping and preservation, bringing down the cost of connectomics through better imaging or cheaper proofreading, lo-fi emulation pathways that predict organism behavior from behavioral and neural data, and shorter experimental feedback loops in neurobiology. (3) Replacement, Enhancement, Biostasis: longevity work including tissue engineering, organ synthesis, 3D bioprinting and progressive replacement strategies; reversible biostasis for organs and full humans with improved preservation and revival protocols and organ banking; and datasets for independently testing genetic predictors, gene editing and better measures of aging. Selected projects receive grant funding, private local compute, workspace at the San Francisco or Berlin AI Nodes, travel-paid field-building events and access to Foresight's advisors and network; in-person contributors at the Nodes are strongly prioritised. Individuals, teams, and nonprofit or for-profit organisations worldwide may apply (for-profits must justify their need for grant funding). Deadline 31 October 2026, 23:59 PDT.
Foresight Institute's AI for Science and Safety Nodes request for proposals funds open, AI-first projects across three calls: I. Local Compute (compute that individuals and communities physically own and control), II. Coordination and Accountability (supercollaboration and decentralised alignment, human empowerment, and AI insurance and open governance: multi-agent oversight and verification, tools for collective deliberation, decision rules for when human control adds value, and independent risk assessment or insurance mechanisms for frontier labs), and III. AI-first Science (automated nanotechnology, whole-brain emulation, frontier bio). Grantees can also get workspace in the San Francisco or Berlin AI Nodes (a community xNode runs in Cape Town) and private compute. All funded code, data and outputs must be open-sourced. Projects are judged on fit with the focus area, impact on reducing existential risk from AI, feasibility within 1 to 3 years, AI-first working methods, ability to execute, and high-risk, high-reward potential. ELIGIBILITY: individuals, teams and organisations worldwide; nonprofits and for-profits welcome (for-profits must justify the need for grant funding). Priority goes to applicants who want to work in person at one of the Nodes, but it is not required. Review takes about three months; applications can opt in to sharing with Lightcone Commons. DEADLINE: 31 October 2026, 23:59 PDT.
The AAAS Science & Technology Policy Fellowships place scientists and engineers inside US federal agencies, congressional offices and the judicial branch for a year of hands-on policymaking, contributing technical knowledge and analytical skills to the federal policy process. AI, data and computing policy placements are among the largest tracks, and AAAS runs a dedicated AI legislative branch cohort that places AI experts in congressional offices at GS-12 Step 10 on the DC locality scale. ELIGIBILITY: US citizens holding a PhD or equivalent doctoral degree in any science or engineering field, or a master's degree in engineering plus three years of post-degree professional experience; some tracks accept a bachelor's plus five years. Relevant expertise includes AI, machine learning, computer science, robotics and data science. Fellows must reside in DC, Maryland or northern Virginia during the fellowship year. TIMELINE: the application portal opens 1 August 2026 and closes 1 November 2026 at 11:59pm ET, with eligibility notification in late November, interview notification January to February 2027, semi-finalist interviews in March, placement offers by early May, and the fellowship year running 1 September 2027 to 31 August 2028. Apply through fellowshipapp.aaas.org.
The Processing Foundation's 2027 fellowship cycle supports artists, educators and creative technologists working on open-source projects, with a USD 10,000 stipend over an 8-month fellowship period. THREE TRACKS: Collective Intelligence (remote, US-based), co-designing AI education tools with youth, aimed at educators with software development or AI/ML experience; Gaming the System (hybrid, Los Angeles), building an open-source toolkit for video game design with underserved communities, for game makers and creative coders based in LA; and DIY Computing (hybrid, New York City), developing open-source hardware and software environments for custom computers, for early-career hardware artists in NYC (applicants currently enrolled in degree programs are ineligible for this track). ELIGIBILITY: US applicants only; hybrid tracks require residency in LA or NYC respectively. International applicants are pointed to the Foundation's Microgrants and Processing Community Day instead. APPLICATION: via nomination form, selecting one track; self-nominations are allowed. Selection by a panel of the Processing Foundation team and external reviewers, judging strength of the project, fit with the Foundation's mission and potential impact. TIMELINE: nominations open 7 October 2026, close 1 November 2026 at 11:59 PM PST; fellows notified 14 December 2026; fellowship runs 25 January to 26 September 2027.
A three-month residency for AI safety generalists: builders, operators and organizers rather than researchers. Residents pitch, build and ship their own projects with weekly advice from field leaders, then get support landing a full-time AI safety role. ELIGIBILITY: open worldwide to early- and mid-career people from operations, communications, policy, engineering, computer science and similar fields; the main criteria are a track record of execution and a genuine commitment to AI safety. TIMELINE: applications close end of day 1 November 2026 (Anywhere on Earth); international applicants are strongly encouraged to apply by 26 October for visa time. Work tests by 9 November, interviews mid-November, decisions by 25 November. Free-response answers must be written without AI assistance.
The NYC track of Eyebeam's AGENCY residency cycle supports New York City-based artists exploring questions of agency at a moment when AI systems increasingly act, decide and carry out tasks on our behalf: how to preserve human agency as these systems shape human behavior, how to assert agency in imagining and building the societies we inhabit, and what agency becomes when understood as collective, relational, entangled and more-than-human. Projects might engage AI directly but do not have to; critical and creative approaches to power, autonomy, technological systems, collective action, human and nonhuman relationships, and the social and material structures of agency are all welcome, including reimagining AI through collective governance, repair, rehearsal or mischief, and repurposing AI systems toward new tools, methods and models for the public good. Residents receive a USD 5,000 stipend, mentorship, and access to tools and technical expertise through Eyebeam's network. ELIGIBILITY: individuals only (no collectives this cycle); not enrolled in a full-time degree program; not committed to another residency during the term; must live in New York City within the five boroughs and be able to work on site at Eyebeam at least one day a week for the full term. TIMELINE: deadline Sunday 15 November 2026, 11:59 PM ET; interviews 7-10 December; notifications 14 December; residency 15 March to 11 June 2027. Info session Monday 19 October 2026, 1 PM ET. Apply via Submittable, free.
Molly Crockett's lab at Princeton, based in the Department of Psychology and the University Center for Human Values, is hiring a postdoc to start in summer or fall 2027. The research topic is flexible. The lab uses methods from computational social science and cognitive science informed by theory from science and technology studies and feminist epistemology, studying feedback loops between individual psychology, institutions and culture, and how the tech industry co-opts these processes to concentrate power. It is especially interested in bridging quantitative and qualitative methods for understanding human experience, and in how culture and power shape which questions are valued and which methods are legitimised. The lab is industry-independent and accepts no research funding from the tech industry; it uses a variety of ML methods and aims to develop thoughtful approaches to automation in scholarship, so it suits researchers excited about AI in science but wary of hype and industry influence. WHAT YOU GET: an USD 80,000 salary from stable institutional funds, initial one-year appointment renewable up to three years. APPLICATION: via Interfolio; questions about fit to mj.crockett@princeton.edu. Review of applications begins 20 November 2026 and continues until the position is filled.
A year-long fellowship for journalists covering artificial intelligence. Fellows complete a three-month full-time remote programme on AI and journalism fundamentals, attend a weeklong summit, then work for nine months as reporters embedded in a partner newsroom (current partners are in Hong Kong, London, Los Angeles, New York, San Francisco, Seattle, Washington DC and Cambridge). ELIGIBILITY: journalism or AI expertise, aged 18+ by 7 June 2027, no degree required; primarily for applicants with the right to work in the US, UK or Hong Kong, as visa sponsorship is uncommon and needs partner-outlet approval.
CISAC's 2027-2028 fellowship application is open across seven tracks: Artificial Intelligence and Laws of Armed Conflict; AI in Strategic Stability and Military Competition; Geopolitics of Artificial Intelligence; India-U.S. Security Studies; Natural Sciences or Engineering International Security (for natural scientists and engineers researching scientific and technical aspects of security topics); Nuclear Security; and International Security. Applicants are considered for all fellowships they are eligible for. Fellows spend the academic year in residence at Stanford engaged in research and writing, participating in seminars and collaborating with CISAC faculty and the wider community. CAREER STAGES: predoctoral (currently enrolled PhD students), postdoctoral (completed PhD or DPhil), and junior faculty or professionals. FUNDING: stipends are provided (predoctoral comparable to Stanford Graduate Fellowships plus tuition; postdoctoral and professional stipends commensurate with experience), though exact amounts are not published; non-US citizens may apply and visas are available through the fellowship. APPLY via the SlideRoom portal (Stanford CISAC Fellowship Program) with the application form, CV, graduate transcript (required for pre- and postdoctoral applicants), research prospectus (max 1,500 words), abstract (max 200 words), writing sample (max 50 pages), and three confidential recommendation letters uploaded directly by recommenders; all materials in English. DEADLINE: 2 December 2026. Questions: cisacfellowship@stanford.edu.
The Institute for Law & AI's US Summer Fellowship is a paid research fellowship at the leading edge of US AI law and policy for law students, professionals, and academics, including the Summer Institute on Law and AI in Washington DC and mentorship from LawAI research staff. FUNDING: USD 2,000 per week, a USD 10,000 relocation/housing stipend for DC-based fellows, covered flights, accommodation and meals for the Convening Week, and potentially higher subsidies for exceptional candidates. ELIGIBILITY: a strong grasp of US legal principles applicable to AI regulation via formal US legal study (incoming 1L and up) or comparable professional experience; full-time commitment 14 June to 20 August 2027; valid US work authorization required (visa sponsorship possible but not guaranteed). The application takes about 20 minutes. DEADLINE: 29 January 2027, 11:59PM ET.
Robot foundation models (pi-zero, OpenVLA, GR00T and others) are approaching widespread deployment on the back of more than USD 18 billion in funding, and almost nobody is working on their safety. The Physical AI Safety Institute, a new nonprofit founded by Bear Haon and Kaylene Stocking and seeded by a grant from BlueDot Impact, describes itself as a capacity-building organisation growing the community that develops techniques to interpret, align and control robot foundation models. Its research fellowship runs in 12-week cohorts: each fellow works within a stream led by a mentor on a concrete problem in physical AI safety, with mentors drawn from researchers with track records in robot foundation models, mechanistic interpretability, safe control or adversarial robustness for embodied systems. Each cohort ends with a research symposium where fellows present their work. STREAMS: mechanistic interpretability (sparse autoencoders and activation patching for VLA and WAM architectures); adversarial robustness (jailbreaking and prompt injection in the embodied action space); runtime monitoring (detecting when a policy operates outside its competence envelope); sim-to-real safety transfer (do safety behaviours survive deployment on physical hardware); and robot security (cyberattack vectors, data poisoning and supply chain for deployed robot models). WHO SHOULD APPLY: researchers with strong technical foundations in robotics or machine learning and a demonstrated interest in safety; prior safety experience is not required. PROCESS: this stage is an expression of interest, not a full application. The form takes about two minutes and asks you to pick your stream and preferred cohort dates; concrete details including stipend figures and application deadlines are shared through that channel, and neither is published on the site as of listing. PAISI also publishes seven free curriculum modules running from classical robot safety through VLA interpretability, adversarial robustness and runtime monitoring to deployment evaluation, plus a white paper, The Case for Physical AI Safety, at paisi.ai/the-case-for-physical-ai-safety. Contact: bear@paisi.ai.
The CHAI Research Fellowship trains highly qualified postdoctoral researchers to advance beneficial AI, working with CHAI faculty Stuart Russell, Pieter Abbeel and Anca Dragan and collaborating with affiliate faculty at Berkeley AI Research and beyond. Fellows have broad freedom to pursue novel research toward provably beneficial AI systems in areas such as reasoning, decision making, learning, multi-agent systems and philosophical foundations, using probability theory, game theory and control theory. Current topics of interest include theoretical foundations of intelligent agent architecture, expressive formal languages for probability models (including probabilistic programming), decision making over long time scales, cumulative lifelong learning, human-in-the-loop planning and reinforcement learning, human-robot interaction, the structure of human preferences, acting on behalf of multiple humans, robust cooperation in heterogeneous multi-agent systems, and mechanism design for human-machine systems. Fellows may also lead collaborations, advise junior researchers, and teach. QUALIFICATIONS: a PhD (or about to be obtained) in a relevant technical discipline (computer science, statistics, mathematics, or theoretical economics) and a record of high-quality published research; prior work on the AI control problem is not required. International applicants are eligible (visa sponsorship provided). TO APPLY: submit via the CHAI application form a resume, a one-page statement of interest describing the research you would like to undertake, and the names and emails of two academic referees. Start date is flexible and applications are reviewed on a rolling basis (no fixed deadline). General enquiries: chai-info@berkeley.edu.
The ILIAD Intensive is a four-week, full-time, in-person program for mathematically strong researchers interested in foundational AI alignment, running in London (at LISA) and in Berkeley. It runs five days a week, 10am to 6pm, in cohorts of 15 to 30 participants. CURRICULUM, across five clusters: Alignment (decomposing the alignment problem, pretraining, post-training and deployment, reward learning); Learning (deep learning theory, singular learning theory, training dynamics); Abstractions, Representations and Interpretability (mechanistic interpretability, natural latents); Agency (reinforcement learning, decision theory, agent foundations); and Safety Guarantees (debate, steganography, the limits of interpretability). MONEY: no tuition is charged. Each participant receives a USD 5,000 travel and housing allowance, office space at the venue, and lunch and dinner five days a week. Participants book and pay for their own travel and accommodation out of the allowance. ELIGIBILITY: a strong mathematical background, typically graduate-level study or a completed degree in mathematics, physics or theoretical computer science, though mathematical maturity is weighted more heavily than specific credentials. Research experience, general competence and motivation are also assessed. Prior machine learning or alignment experience is not required. DEADLINE: no fixed closing date is published. A single application covers several upcoming programs and applicants indicate their preferred dates and locations; applications are being accepted for the October 2026 and November 2026 cohorts. HOW TO APPLY: through the application portal at iliad.ac/apply. Questions to contact@iliad.ac.
The Anthropic Fellows Program is a four-month, fully funded research fellowship with direct mentorship from Anthropic researchers. Fellows work on empirical projects aligned with Anthropic's research priorities using external infrastructure, aiming to produce public outputs such as papers. Five workstreams: AI Safety, AI Security, ML Systems and Performance, Reinforcement Learning, and Economics and Policy; safety topics include scalable oversight, adversarial robustness, mechanistic interpretability, and model organisms. FUNDING: a weekly stipend of USD 3,850 (GBP 2,310 / CAD 4,300), around USD 15,000 per month in compute funding, and other research expenses. ELIGIBILITY: fluent Python; available full-time for four months; strong technical background in computer science, mathematics, or physics; work authorization in the US, UK, or Canada (no visa sponsorship). Locations: San Francisco, London, Ontario, or remote within those countries. APPLICATION: rolling, via the Constellation portal; the process includes an initial application, reference checks, technical assessments, interviews, and a research discussion. The next cohort is expected to start in January 2027, so apply well before the end of 2026.
NOTE: this tier is institutions-only; individual researchers cannot apply in a personal capacity but can serve as principal investigators through their institution. Anthropic's earlier Economic Futures Research Awards (USD 10,000-50,000 plus USD 5,000 Claude API credits for empirical research on AI's economic impacts, rolling, with results presented within six months) are currently not accepting applications; this entry covers the larger Research Fund now open for proposals. THE FUND: Anthropic has committed USD 200 million to the Economic Futures Research Fund to support ambitious external research on interventions that prepare society for the economic impacts of AI, studying what programs could make the economy more flexible and resilient, ensure AI's benefits are shared, and minimise harm from AI-driven disruption. FIVE RESEARCH PRIORITIES: (1) AI's impact on workers at the firm and workplace level; (2) equipping people for AI-driven transitions; (3) modernising income support for displacement; (4) building worker stakes in AI-driven growth; (5) generating evidence on public investments. FUNDING: projects primarily in the USD 5-30 million range, minimum USD 1 million, flexible upward for well-scoped, high-potential-impact projects. ELIGIBILITY: accredited universities and other degree-granting institutions, independent research institutes and policy research organisations, and nonprofits with a track record of running field experiments at scale; a global fund with an initially US-centric focus. APPLICATION: via the request for proposals linked from the research agenda announcement (Google Form); no deadline published as of September 2026.
Mentored, individual research in applied mathematics for AI alignment, including foundational questions that get less attention elsewhere, drawing on mathematics, theoretical physics and computer science. ELIGIBILITY: designed for PhD students, postdocs and researchers with equivalent experience in mathematics, theoretical physics, computer science or related fields; selection weighs academic credentials, technical ability, intentions and fit with the cohort's research coverage. APPLICATION: rolling, with fellowships starting at the beginning of most months.