AI Fellowship Rejection Letter Templates and Lessons

This post is based on data and insights gained from rejection letters received from a number of top organizations, which I have compiled and analyzed in this post.

Introduction

Every organization that runs fellowships, courses and programs in AI safety in which they receive applications should have at least two rejection letters which have meaningful next steps for applicants to help them upskill, get ready to apply for the next cohort or redirect them to a program that better suits their current level or area of expertise. This is a relatively low effort thing that can be done to make sure applicants are not completely dejected and immediately fall out of the AI safety space. It won’t perfectly solve drop-out but it can serve as another tool in field-building to maintain interest and encourage upskilling.

The two bare minimum rejection templates that I recommend are one for Near Misses (people who just barely missed the cut-off for the program) and Long-Shots (people who need significant upskilling). This would require that those receiving applications also carefully review applications to identify who these Near Misses are. 

Ideally, Near Miss rejection letters should be tailored to individual cases and separated based on specific reasons for the rejection. Individualized feedback on why the applicant didn’t make it are the best option, but not always an option for organizations that struggle with capacity constraints. 

A few key features that I’ve observed in good rejection letters are:

• A warm letdown 
• The ability to be added to a mailing list or Substack where upcoming opportunities are sent from
• A catalogue of other programs and courses that the applicant can take to upskill. These should match the level the applicant is at (for example, an introductory BlueDot course is likely insufficient for someone applying to MATS)
• Other ways to get involved with the organization, if they exist
• Reasons that the applicant was rejected and specific qualities the organization looks for in their accepted applicants
• A good understanding of the “ladder” or “grid” of different organizations and which ones might be relevant and a better starting point based on the program they just applied to

I am going to include here two rejection letters that I made with the help of Claude based on the compilation of rejection letters from 26 different organizations. These are the rejection letters I’ll be using for my own fellowship at the Lisbon AI Safety Hub. Anyone is free to copy these verbatim and adapt them as they please for their own rejection letters. If you are an organization who currently sends rejection letters with no next steps or follow-up information, I strongly encourage you to do so. 

Long-Shot Application: More general for those who clearly didn’t fit with our program. 

Hi there,

Thank you for applying to the Foundations of AI Safety Fellowship at the Lisbon AI Safety Hub, and for the thought you put into your application.

We had many more strong applicants than places in our first Lisbon cohort, so we're not able to offer you a spot this time. Please don't read too much into it — with a small classroom and a competitive first round, we've had to turn away people we'd have been glad to work with. It's not a judgement on your potential to contribute to AI safety.

The good news is that almost everything we cover, you can start on your own right now — and we'd love to keep you in our orbit while you do.

Where to start:

• Newer to the field? Work through BlueDot's AI Safety Fundamentals or read the AI Safety Atlas at your own pace.
• Already have the basics? The best next step is hands-on: apply to a mentored research programme like SPAR, or try an Apart Research weekend hackathon. You'll find these and much more on the field map on the AI Safety Map.
• Thinking about direction? 80,000 Hours has a free AI-safety career guide and 1-1 advising.

Stay connected:

Everything runs through our site. Join our Discord and subscribe to our Substack there, and you'll be first to hear when applications open for our next cohort, plus invitations to our monthly Lisbon meetups and free online guest lectures. You're welcome at all of them whether or not you're in the fellowship.

If you’ll be in Lisbon and you want to get involved in the Hub in the future, we'd genuinely like to see you apply again for Cohort 02. Thank you again for your interest in what we're building in Lisbon.

Warmly,

The Lisbon AI Safety Hub team

www.lisbonaisafetyhub.org

Near Miss Application: Those who demonstrated strong skills but we couldn’t take for one reason or another

Hi Caleb

Thank you for applying to the Foundations of AI Safety Fellowship at the Lisbon AI Safety Hub. We're writing with a genuine mix of disappointment and encouragement: you were among the strongest applicants to our first cohort, but we lack the capacity to have you in this cohort. Rather than just an outright rejection, we are writing this to show that we want you to remain a part of LAISH and the AI safety community in general because we think you have something to contribute. 

Since we'd like you to stay close to LAISH, it's worth being open about what we weight most heavily when we choose who to work with:

• Being based in (or moving to) Lisbon, and looking like someone who'll contribute to the Hub well beyond the fellowship itself.
• A clear-eyed grasp of both the capabilities and the risks of frontier AI.
• A real openness to steering your studies or career toward AI safety.
• Genuine alignment with LAISH's mission and the values of the wider AI safety community.

You stood out on some of these, which is exactly why we don't want to lose touch. 

A few concrete ways to stay involved:

Our first hackathon. We're running a LAISH hackathon soon [dates + details on our site], and we think you'd be an excellent fit. This is a chance to turn an idea into a real project over a weekend, meet collaborators, and produce something you can point to. We'd love for you to apply; sign-up is on our site and in our Discord.

Stay in our orbit. Join our Discord and subscribe to our Substack. Make sure to also closely follow our website. That's where we announce the hackathon, Cohort 02, our monthly Lisbon meetups, and free online guest lectures, all of which you're welcome to join.

Since you've most likely worked through the introductory courses already, here's where we'd point someone at your level next:

• Get hands-on. Mentored, project-based programmes are the best next step — SPAR (part-time, remote), ARENA (technical), Gen Stream (generalist), and Future Impact Group. Apart Research runs weekend sprints that pair perfectly with our hackathon.
• Build something and publish it. A small, finished project — an eval, a replication, a focused write-up on LessWrong or the Alignment Forum — is the single strongest signal for your next application. It beats taking another course.
• Go deep and build research taste. Follow current work from the Alignment Forum and from lab teams (Anthropic, Google DeepMind, Redwood, METR), and keep up with a newsletter or two — Transformer and the AI Safety Newsletter are good, as are Don’t Worry About the Vase and ChinaTalk for slightly more niche topics.
• Find the next opportunity. AISafety.com's training list and field map track open fellowships and events as they appear.

And please do apply again for Cohort 02 — we'd be glad to have you. In the meantime, jump into the Discord, come to a meetup, and put the hackathon on your calendar.

Warmly,

The Lisbon AI Safety Hub team

www.lisbonaisafetyhub.org

I am open to advice on how to keep improving these. If you have more examples of rejection letters that were not covered in my previous post, please send them to me so I can update the data and continue to improve rejection letters and help rejected applicants upskill: [email protected]