AI Alignment Foundation Research Fellowship

AI Alignment Foundation Research Fellowship is sponsored by AI Alignment Foundation. This fellowship supports individuals interested in researching how increasingly capable artificial intelligence systems can be developed and deployed safely.

Fellowship | AI Alignment Foundation Eight weeks doing real alignment research. A paid, full-time research fellowship for AI alignment researchers, run by AIAF in partnership with AE Studio. You'll work at the speed of an industry research team, experiencing our agile methodology to accelerate research, on projects that actually solve alignment.

Apply Now See how it works Applications are now open · Program starts September 8, 2026 Research, run at industry speed. You'll have dedicated support throughout, interfacing with several stakeholders who will support your work. Here's how it's structured: Your research manager: day-to-day support A dedicated research manager works alongside your team every day.

They help with weekly prioritization, experiment tracking, and notes, and keep things moving. They will be your closest point of contact throughout the program. AE researchers: set the vision AE researchers act as visionaries for your research.

You'll discuss your progress with them, get project direction, and respond to their feedback. Midpoint check-in: senior mentorship Halfway through, you'll have a one-on-one with someone from our leadership or our partner AE Studio's alignment leadership, to give guidance on your research, your career prospects, and feedback on your work. Along the way, each team works toward a LessWrong post and, when the research supports it, a paper.

The program closes with a Demo Day presentation. Full-time, work from anywhere For the full 8-week program Cloud + API budget per team In-person option with travel covered Highly selective · reviewed individually The problems we're working on. Fellows work on projects defined by AIAF, in collaboration with AE Studio's research team.

Specific assignments happen after selection, based on background fit and the cohort's strengths. Reading about our current directions will give you a sense of the kind of research we do. 01 Modular Pretraining Enables Access Control 02 Self-Interpretation in Language Models via Adapter Probes 03 Steering Resistance: Self-Correction Circuits in Large Language Models See what else we're working on Strong candidates look like this.

Strong ML background. You can run experiments, implement papers, and debug your own code. You've spent real time thinking about AI alignment, not just reading about it.

You've done some form of independent research before, a LessWrong post, a hackathon project, a thesis, a preprint, anything that shows you can take initiative and produce something. You can work autonomously for a week and show meaningful progress at the end of it. Any career stage: recent PhDs, MSc students, self-taught researchers, or people transitioning into AI alignment from another field.

This probably isn't for you if... You need significant hand-holding to get started on a new problem. You've never written code to run an experiment.

You're applying for the credential but aren't genuinely interested in alignment long-term. You have a specific research agenda you want to pursue. We're looking for people who are flexible and excited about exploring different directions.

You want to work on policy or governance-focused alignment. This fellowship is for technical research. For the full 8-week program, paid in installments.

Cloud GPU access and API credits per team. No out-of-pocket costs for running experiments. Attend in person in LA with all travel and expenses covered, or join remotely.

We can't sponsor visas, but if you can make it, we'll handle the rest. Linear, GitHub, Google Drive, Slack, plus AIAF's full research methodology and experiment standards. Every team works toward a LessWrong post and reproducible work.

We'll work closely with your team on the opportunity to publish a paper too. AIAF is committed to helping fellows find their next step. We work with our partner organizations, including AE Studio, to connect strong fellows with opportunities that fit.

Timeline and what we need. CV and a short questionnaire covering your background, what you're currently working on, and links to any relevant work (GitHub, papers, LessWrong posts, projects). Plus a short video.

Short video 5 minutes max Recorded on whatever device you have, no production setup needed. You'll need to answer two questions: 1 . Why does AI alignment matter to you?

2 . After reading about our research , pick one direction that interests you. Briefly describe at least one critique you have of it, something you would extend, and what you think could be most promising to explore next.

Technical test Shortlisted candidates only If we like your application, you'll be invited to complete a short technical test. A conversation covering your research background, alignment thinking, and fit with the cohort. Applications are now open.

Do I need to know which research area I want to work on before applying? No. We assign projects after selection, based on background fit and the cohort's strengths. Is there a hiring guarantee at the end?

No. There's no guarantee. What we can say is that AIAF is committed to helping strong fellows find their next step, and we actively work with our partner organizations to connect people with opportunities that fit. Is the fellowship fully remote?

Yes. The eight weeks are fully remote. For Demo Day at the end of the program, we're planning a hybrid event in LA.

If you can make it in person, we'll cover all travel and accommodation. If you can't, you'll be able to join and present remotely. Please note we're not able to sponsor visas, so attending in person requires that you sort out any travel documentation yourself.

Can I apply if I'm not based in the US? Yes. The fellowship is fully remote, so location isn't a barrier.

But you need to be open to meetings at times that overlap with other stakeholders, for example mornings Pacific Time. What if I'm still finishing a PhD or master's? Fine, as long as you can commit to 40 hours per week for the full 8 weeks.

Be honest with yourself about your availability before applying. I'm mid-career and transitioning into AI alignment from another field. Should I still apply?

Yes. We're looking for strong technical ability and genuine engagement with alignment, not a specific career stage. People moving in from adjacent fields are very welcome.

What does the review process look like? We review every application. If we like what we see, you'll move to a short technical test.

Candidates who pass that are invited to a final interview. We aim to send decisions by early September. The $12,000 stipend is paid in 2 installments over the duration of the program.

If anything here is unclear, reach out and we'll get back to you. Applications are now open. Deadline August 17, 2026.