Opportunities in AI Safety & Governance
There are many fellowships, courses, and programmes which can help you get experience in mitigating risks from advanced AI systems. We really need all hands on deck to steer the rapidly advancing capabilities of AI in safe and beneficial directions, and there are (sadly!) only so many individuals we can support with our finite resources at ERA. We care deeply about our mission, and in that spirit, we have made a list of some other opportunities that you might be interested in.
1. BlueDot Impact – Various Courses (Online): 2
2. Alignment Research Engineer Accelerator (ARENA) (London, UK): 2
3. Center for AI Safety – ML Safety Course (Online): 3
Our recommendations for domain upskilling in ML 5
Technical (& Technical Governance) Fellowships 6
1. ML Alignment Theory Scholars (MATS) (Berkeley, CA & global): 6
2. Pivotal Research Fellowship (London, UK): 6
3. Supervised Program for Alignment Research (SPAR) (Remote): 7
4. Principles of Intelligent Behavior in Biological and Social Systems (PIBBS) (US or UK) 7
5. London AI Safety Research Labs (LASR Labs) (London, UK): 7
6. Apart Labs Fellowship (Remote): 8
7. Impact Academy Global AI Safety Fellowship (Remote) 8
8. Algoverse AI Safety Fellowship (Remote): 9
9. Anthropic Fellows Program (Remote) 9
10. OpenAI Safety Fellowship (Berkeley, remote) 9
Policy & Governance Fellowships 10
1. RAND Technology and Security Policy Fellowship (US/UK – Remote or In-Person): 10
2. Horizon Fellowship (Washington, DC) 10
3. Institute for AI Policy & Strategy (IAPS) Fellowship (Washington, D.C. or Remote): 11
4. Law & AI Institute Summer Fellowships (Remote-first with US & EU options): 11
5. GovAI Summer/Winter Fellowships (London, UK): 12
6. Talos Fellowship (Brussels, Remote) 12
7. Tarbell Fellowship (Various) 13
8. Generator Residency (Berkeley) 13
AISafety.com
http://AISafety.com is a hub for AI safety resources, including organisations, fellowships, events & training as well as options for self-study, as well as many other resources. Take a look at their map for an interactive overview!
Courses & Bootcamps
BlueDot Impact – Various Courses (Online):
BlueDot Impact offers a series of free courses which introduce key concepts in AI safety, alignment, and governance. These multi-week courses combine curated readings, lectures, and discussion groups to help you build a foundation and network in AI safety. Tracks include an AGI Strategy, Biosecurity, Governance and Technical AI Safety Course.
Target: Students and professionals new to AI safety– no technical background is required for the intro courses. It’s designed as an on-ramp for anyone curious about AI’s global risks and wanting to contribute.
Format: Online, with options for self-paced learning or cohort-based discussion groups led by facilitators. (BlueDot’s flagship “Future of AI” intro is self-paced ~2 hours, while the deeper Strategy, Biosecurity, Governance and Technical courses run 5 weeks with weekly seminars.)
Application: Open and recurring – Many courses have open registration (no application) or run multiple cohorts per year. Check BlueDot’s Courses site for current offerings and schedules.
Alignment Research Engineer Accelerator (ARENA) (London, UK):
An intensive 4–5 week bootcamp designed to upskill strong programmers in machine learning engineering for AI alignment roles. Participants receive training in applied ML, alignment problem-solving, and career mentorship, with the aim to place them in technical alignment positions.
The Entire Curriculum is available here for self-studying and upskilling!
Target: Early-career software engineers or CS students who care about AI safety – requires solid Python skills and math background.
Format: In-person in London for the duration (travel, accommodation, and meals are fully covered). It’s a short, immersive cohort program with hands-on projects and close mentorship.
Application: Not currently open – future iterations may be announced (interested candidates should monitor the ARENA website and fill in the expression of interest form).
Center for AI Safety – ML Safety Course (Online):
An 8-week virtual course focused on machine learning safety research, created by Dan Hendrycks and the Center for AI Safety. It provides a structured curriculum covering robustness, monitoring, alignment, and systemic safety in ML systems.
Target: Students or professionals with a basic deep learning/ML background who want to transition into AI safety research. (It’s recommended for those who have taken an intro ML course; it does not overlap much with conceptual AGI fundamentals, focusing instead on empirical safety research.)
Format: Online cohort, ~5 hours/week commitment for 8 weeks.
Application: No application – self-paced course.
ML4Good Bootcamps
An intensive 10-day in-person bootcamp that provides comprehensive training in AI safety fundamentals. The curriculum combines technical skill development with exposure to governance, strategy, and risk assessment. Participants engage in peer-coding sessions, expert presentations, literature reviews, and a 2.5-day capstone project.
Target: Individuals with technical backgrounds interested in contributing to AI safety work - prior programming experience and basic ML knowledge recommended.
Format: In-person immersive cohort experience (with all expenses typically covered). The program is regularly updated to reflect rapid developments in AI safety.
Application: Periodic offerings – no currently scheduled bootcamps, but interested candidates could join their mailing list for announcements about future cohorts.
Our recommendations for domain upskilling in ML
Intro Level ML Courses (1 of these):
Subject-focused courses (at least 1, ideally all 3):
- Computer Vision: UMich: Deep Learning for Computer Vision
- NLP: Stanford: Natural Language Processing with Deep Learning
- Reinforcement Learning
Understanding-based courses
(listed in decreasing order of importance, ideally all 3, but the first one is must and very relevant to LLMs!)
- Deep Multi-Task and Meta Learning (see recordings of previous lectures)
- Deep Unsupervised Learning
- Optimisation Algorithms or Boyd's Convex Optimisation
Other courses & a paper
Very interesting and useful (especially for some types of ‘safety’/understanding neural nets research):
- Roger Grosse’s courses, especially AI Alignment and Neural Network Training Dynamics
- Deep Learning Theory: thesis by Preetum Nakkiran, particularly section 1.3.2
- Recent paper led by Usman Anwar, an ERA research manager from 2025, helpful to anyone trying to get into AI Safety
Research Agendas
We think these research agendas have several shovel-ready project directions that you could investigate in your own time to upskill & contribute meaningfully to the field:
Technical AI Safety
- Foundational Challenges in Assuring Alignment and Safety of Large Language Models
- Open Problems in Mechanistic Interpretability
AI Governance
Technical AI Governance
Technical (& Technical Governance) Fellowships
ML Alignment Theory Scholars (MATS) (Berkeley, CA & global):
A 10-week intensive research scholars program that pairs selected participants with leading AI alignment researchers for mentorship. Scholars conduct independent or joint research on alignment theory or interpretability, attend seminars and workshops, and connect with the Berkeley AI safety research community.
Target: Talented individuals with a strong grasp of the AI alignment landscape (e.g. AGI Safety Fundamentals alumni) and research experience (often postgraduates or PhD-level).
Format: In-person seminars (Berkeley hub) with some support for remote scholars; no direct funding from the program, but AI Safety Support offers a stipend (~$12k for 10 weeks) to participants.
Pivotal Research Fellowship (London, UK):
A 9-week in-person summer research fellowship (formerly CHERI) hosted at the London Initiative for Safe AI, focusing on global catastrophic risk reduction research. Fellows undertake full-time projects in technical AI safety, AI governance, or biosecurity, mentored by experienced researchers and supported by research managers.
Target: Early-career individuals (students, graduates, or professionals) with strong interest in AI risk research (even with limited experience).
Format: In-person cohort in London (includes a stipend ~£5,000 and travel + accommodation support), with office space and meals provided.
Application: Annual summer & winter fellowships; check for next year’s cycle if missed.
Supervised Program for Alignment Research (SPAR) (Remote):
A 3-month virtual mentorship program connecting small teams of mentees with expert mentors to tackle AI alignment or governance research projects. SPAR runs part-time, allowing students or professionals to contribute ~10 hours/week to a supervised research project.
Target: Early-career individuals with either a technical/STEM background or a policy/social science background – projects span technical safety and AI governance topics.
Format: Remote, volunteer basis; includes regular mentor meetings, team collaboration, and project deliverables.
Application: Recurring cycles (e.g. Spring and Summer). Keep an eye out for announcements.
Fellows work on selected projects at the intersection between their field of expertise and AI safety. Fellows will work in close collaboration with a mentor who will help them effectively navigate the AI Risk landscape and apply their knowledge to it.
Target: Ph.D. and Postdoctoral researchers (or comparable research experience) from various fields, mostly studying complex and intelligent behavior in natural and social systems but also those studying mathematics, philosophy or engineering, who are motivated by the mission of making AI systems safe and beneficial.
Format: Hybrid, including remote 8-week reading group and in-person retreat and office co-working. Fellows receive a stipend of 3,000 USD/month and are expected to work full-time on their projects over the course of the fellowship.
London AI Safety Research Labs (LASR Labs) (London, UK):
A 13-week in-person research program where participants work in teams of 3–4, each team supervised by an experienced AI safety researcher. The goal is to “learn by doing” – each team proposes, executes, and writes up a research project aimed at preventing loss of control of advanced AI.
Target: Technical AI safety researchers (often those looking to join alignment labs or partway through their PhD) with solid ML/CS background.
Format: Full-time, in-person in London (at the Safe AI Institute offices), with a generous stipend (~£11,000 for 3 months) and travel, food, and office space provided. Participants start with a workshop week to choose projects, then spend 12 weeks on research and paper writing.
Apart Labs Fellowship (Remote):
A 3–6 month remote research incubator which helps participants turn AI safety project ideas (often from hackathons) into publishable research.
Target: Aspiring AI safety researchers (often professionals or grad students) looking to transition into the field, with solid technical skills demonstrated through Apart’s research sprints.
Format: Fully remote mentorship, flexible pacing; teams receive expert guidance, GPU resources, and project management support.
Application: Ongoing – fellows are invited based on performance in Apart Research Sprints (no fixed annual deadline).
A 3–6 month fully-funded research fellowship which places exceptional STEM talent at leading AI safety organizations. Fellows work on advancing safe and beneficial AI development, undergoing an initial bootcamp with personalized mentoring and then a research placement with partner orgs like CHAI (UC Berkeley), Conjecture, FAR AI, or the UK AI Safety Institute.
Target: Strong programmers or researchers worldwide (must be 18+) – e.g. those with ML research experience or industry software backgrounds – looking to transition into AI safety research.
Format: Full-time; in-person preferred at host organizations (with hybrid/remote possible case-by-case).
Algoverse AI Safety Fellowship (Remote):
A 12-week online program designed to help university students and professionals transition into AI safety research. Participants explore various AI safety research agendas to find their ideal fit in the field, working in teams of 3 under mentorship from researchers at leading organizations.
Target: University students and professionals with strong programming and AI/ML experience who need guidance navigating the AI safety field.
Format: Fully remote, part-time commitment (25+ hours per week), completely FREE for participants. Teams of 3 are paired with mentors from Apollo Research, FAR.AI, MATS, University of Washington, and Oxford University. The objective is to publish novel research at EMNLP/NeurIPS workshops or the EACL main conference.
Anthropic Fellows Program (Remote)
A 6‑month, fully-funded AI safety research fellowship run by Anthropic, where 10–15 fellows work full-time on projects (e.g. adversarial robustness, scalable oversight) under mentorship from Anthropic researchers.
Target: Engineers and researchers (at various experience levels) looking to pivot into AI safety research.
Format: Full-time remote collaboration (with optional co-working space in San Francisco or London).
OpenAI Safety Fellowship (Berkeley, remote)
A hybrid program running from September 14, 2026 through February 5, 2027 where fellows will work closely with OpenAI mentors and engage with a cohort of peers. Priority areas include safety evaluation, ethics, robustness, scalable mitigations, privacy-preserving safety methods, agentic oversight, and high-severity misuse domains, among others. The fellowship includes a monthly stipend, compute support, and ongoing mentorship.
Target: Open to applicants from a range of backgrounds, including computer science, social science, cybersecurity, privacy, HCI, and related fields with a priority for research ability, technical judgment, and execution over specific credentials. Reference contacts will be required.
Format: Fellowship will be based out of the Constellation workspace in Berkeley, though fellows may also work remotely. Applications close on May 3, 2026.
Policy & Governance Fellowships
RAND Technology and Security Policy Fellowship (US/UK – Remote or In-Person):
A 1–3 year fellowship at RAND Corporation for independent research at the intersection of emerging technology and security policy. Fellows pursue self-directed research on topics like AI governance, receive mentorship from RAND experts, and may contribute to RAND projects one day per week.
Target: Aspiring policy analysts from undergraduate to mid-career level – a wide range of experience is accepted. (Many fellows have graduate degrees or policy experience, but it’s open to exceptional undergrads as well.)
Format: Flexible – can be full-time or part-time; remote or at a RAND office in the US or UK (Washington D.C., Santa Monica, Pittsburgh, Boston, or Cambridge UK). Comes with a generous stipend/salary (scaled by experience; e.g. ~$40k–$200k/year full-time in the US) and health benefits.
Application: Rolling admissions (open now) – no hard deadline (applications reviewed quarterly). Fellows must be eligible for a security clearance in US/UK, so citizenship/residency requirements apply.
Horizon Fellowship (Washington, DC)
A full-time public service fellowship placing fellows in U.S. government and policy roles to tackle emerging technology issues, especially AI governance. Fellows receive policy training, mentorship, and are matched with a host office in the Washington, DC area (e.g. federal agencies, Congressional offices, or think tanks) for 6–24 months.
Target: Early-career and mid-career professionals with expertise in AI or biotech who seek to transition into government/policy work (requires U.S. work authorization).
Format: In-person in Washington, DC; one-year placements (renewable to two) for government tracks, or 6-month renewable terms for junior think-tank track.
Institute for AI Policy & Strategy (IAPS) Fellowship (Washington, D.C. or Remote):
A fully-funded 3-month fellowship for professionals to tackle real-world AI governance challenges and develop practical policy skills. Fellows work closely with IAPS mentors on projects focused on frontier AI model governance (e.g. drafting policy memos, analyzing regulations, briefing officials). The program includes weekly 1:1 career coaching and network events to springboard fellows into long-term AI policy roles.
Target: Professionals pivoting into AI policy from diverse backgrounds – government, law, tech, academia, etc.. Two tracks cater to experience level (Fellow vs. Senior Fellow).
Format: Full-time; starts with a required 2-week in-person residency in D.C., then remote work is allowed for the remaining time. Stipends are provided (approx. $15,000 for Fellows and $22,000 for Senior Fellows) for the 3 months.
Law & AI Institute Summer Fellowships (Remote-first with US & EU options):
Paid summer research fellowships for those interested in the legal and policy aspects of AI. The program has two streams:
- US Fellowship: 8–12 week fellowship focusing on US AI law & policy.
- Format: Remote-first, with a 1-week in-person intensive in Washington, D.C. (June 15–Aug 21, applications now closed) for workshops and networking. Fellows receive ~$1,500 per week stipend and mentorship from LawAI researchers
- EU Fellowship: 8–12 week fellowship focusing on EU AI law (e.g. EU AI Act).
- Format: Remote-first, with a 1-week in-person in Brussels or Cambridge, UK for EU-specific policy engagement. Fellows receive about €1,200 per week and travel, accommodation, and meals are covered for the in-person week.
Target: Law students, legal professionals, and academics who want experience at the cutting edge of AI governance, policy, and law. No prior AI-specific experience is required – just strong legal research skills and interest in AI issues. Fellows work on a research project (e.g. law review article or policy memo) under a mentor and get exposure to policymakers and experts.
GovAI Summer/Winter Fellowships (London, UK):
Hosted by the Centre for the Governance of AI, these are 3-month research fellowships in London for early-career talent to work on AI governance research of their choice. Fellows design a policy or governance research project (with guidance), and are paired with a primary mentor from GovAI plus a secondary mentor from the broader network. The program also includes Q&A sessions with AI governance experts and work-in-progress seminars.
Target: Early-career individuals in fields like public policy, political science, law, economics, history, sociology, or computer science who want to specialize in AI governance. (Many have or are pursuing graduate degrees.)
Format: Full-time, in-person in London is encouraged (desk space and weekday lunches provided), though exceptional candidates can sometimes do part-time or remote. Fellows receive a stipend (~£11,000 for 3 months) plus travel support to London.
Application: Periodic – GovAI runs summer and winter cycles.
Talos Fellowship (Brussels, Remote)
Run by the Talos Network, this 7-month programme fast-tracks early-career talent into European AI-policy roles. Fellows first complete an EU AI Policy Fundamentals curriculum (8 weeks, part-time, online) and a 7-day, in-person policymaking summit in Brussels; those on the Placement Track then spend 4–6 months embedded at a leading tech-policy think tank, receiving a living stipend.
Target: Early-career graduates (BA/BSc completed by start-date; many hold master’s degrees) who care about AI governance and can thrive in fast-moving Brussels policy work. EU citizenship is strongly preferred, but outstanding non-EU candidates with ≥3 years’ relevant experience are considered. Backgrounds span public policy, ML/CS, economics, law, or adjacent fields.
Format:
- Training Track (≈3 months, part-time): 9-week online course (5–10 h/week) + 1-week Brussels Summit (travel & accommodation fully covered). Candidates can also apply for up to two months of needs-based job-search funding.
- Placement Track (≈7 months total): includes everything in the training track, plus a 4–6-month full-time placement (usually Brussels-based; other locations include Paris & London). Fellows receive ~€2,000 / month during the placement, with higher stipends available for more experienced candidates.
Tarbell Fellowship (Various)
A prestigious one-year fellowship for early-career journalists committed to covering AI and other emerging technologies in the public interest. The Tarbell program provides a 9-month placement at a major newsroom (such as Wired, The Washington Post, etc.), where fellows work as technology reporters, plus a 3-month period of intensive study and project work.
Target: Aspiring journalists with 0–5 years of experience (or mid-career professionals transitioning into journalism) who want to specialize in AI risk, tech policy, or tech ethics reporting.
Format: Largely in-person/onsite work at the host newsroom (which could be in various global locations depending on placement) with portions of the fellowship (study group, summit) being cohort-based and remote/online. A stipend of $50,000 USD for the year, plus travel expenses for training, is provided.
Generator Residency (Berkeley)
Generator is a 3-month residency where residents pitch, build, and ship projects that build capacity and infrastructure across the AI safety ecosystem, and then get support landing full-time roles at the orgs that need them. Applications for the Summer 2026 cohort close on April 27, 2026.
Target: For those who don’t want to be a ‘researcher’ but are highly agentic, ambitious with a track record of execution and genuine commitment to making AI go well for humanity.
Format: In person at the Constellation office in Berkeley with a monthly stipend of $6k/month + travel and accommodation.
National Security Law and AI (Vista Institute)
This virtual, part-time course is designed for policy professionals, staffers, researchers, academics and recent graduates working on, or moving towards AI policy and the national security challenges posed by advanced artificial intelligence systems. Applications for the 2026 cohort close on July 19, 2026.
Target: People who intend to engage with national security law and AI in their work: recent graduates, doctoral and academic researchers, government staffers, early-career AI policy researchers, and professionals moving into the field.
Format: Virtual from August 10 to 26, 2026.
AI Agents and the Law (Vista Institute)
This virtual, part-time course is for people who want to learn about and help shape how the law applies to increasingly capable AI agents. Applications for the 2026 cohort close on July 19, 2026.
Target: People from a wide range of disciplines: advanced graduate students and professionals from law and computer science, as well as US government professionals, will form the core of this year's cohort. Professionals from other disciplines are welcome to apply so long as they're genuinely intending to work on the novel legal challenges that AI agents create.
Format: Virtual from August 19 to October 2, 2026.
Job & Opportunity Boards
- The 80k job board provides a curated list of publicly advertised vacancies that are particularly promising. They post opportunities, internships, fellowships, and jobs that contribute to solving key global problems. 80k can also help you develop the career capital — skills, experience, knowledge, connections and credentials — to solve these problems in the future through 1-on-1 career guidance.
- They also offer career advising
- Fellowship masterlist, with programs in AI safety, technology policy, cybersecurity, and more.
SuccessIf: Career Advising - AI-Safety focused Career Advising
Database of US Policy Fellowships