About the challenge

Welcome to the first multi-disciplinary summer challenge organized by SAIN Utrecht. You can participate individually or with a team. You will showcase your passion, ideas, and talent by crafting a short study over 6 weeks. The challenge is about shedding light on underexplored areas of AI Safety, replicating existing studies to perform new analysis, or producing surveys to raise awareness inside and outside the field. Just like professional researchers, you will be working under uncertainty. This is your chance to narrow down broad, general questions into a highly actionable study.

The challenge in a nutshell:

You can choose any area of AI Safety, including Technical AISF or Governance. We are providing no strict theme or step-by-step guidelines on how to proceed, racking your brain is the only way to go. You will spend 6 weeks defining your scope, conducting your research, and synthesizing your findings.
Working hard for these 6 weeks is for committed individuals eager to show what they are made of. Therefore, your final submission must be a comprehensive essay published as a Substack post.
You can participate alone or with a team.
Five mentors are available to help you throughout the challenge.
Participant must be based in the Netherlands.

Dates

Monday, 6 July 2026 - Hybrid Kick-off event with Challenge Introduction and Jury Panel Discussion on SAIN Research Hub.
Friday, 10 July 2026 - Team Roster Finalization & Networking Drinks.
Monday, 13 July 2026 - Late sign-up deadline: after this date we no longer accept new participants.
Monday, 17 August 2026, Anywhere on Earth - Substack post submission.

Information on the location for the kick-off event, networking and Winner ceremony can be found on LUMA:

https://luma.com/sain-utrecht-events

Our other channels:

https://linktr.ee/sainutrecht

Requirements

What to Build

Your final submission is a comprehensive essay, published as a Substack post, that presents your or your team's research on your chosen challenge question. The essay should read as a self-contained piece of work: someone outside the challenge should be able to follow your argument, understand your reasoning, and see the evidence behind your conclusions.

Some substack examples:

Opus 4.6 Reasoning Doesn't Verbalize Alignment Faking, but Behavior Persists

Standards, Incentives, and Evidence: The Frontier AI Governance Triad

Length and depth

Aim for a substantial piece (minimum 2,000 words). Go deep on one question rather than broad across several.

What to Submit

Please submit your research by emailing the link to your published Substack on Devpost.

Each submission should include:

The essay itself, published live on Substack
A title
Named contributors and their roles
Any references, data sources, or tools cited clearly
For a more detailed explanation on the output format, check the following document:
Substack Format

Hackathon Sponsors

Prizes

3 non-cash prizes

1st Place

1 winner

Exclusive SAIN Merch - Gold Tier Medal
The opportunity to pitch at our New Academic Year Kick-off Ceremony
Be scouted for a position in the SAIN Research Hub for the next academic year

2nd Place

1 winner

Exclusive SAIN Merch - Silver Tier Medal
Be scouted for a position in the SAIN Research Hub for the next academic year

3rd Place

1 winner

Exclusive SAIN Merch - Bronze Tier Medal
Be scouted for a position in the SAIN Research Hub for the next academic year

Judges

Luca Dughera

Riccardo Campanella

Cem Kaya

Ana Paula Castillo Rodriguez
SAIN

Judging Criteria

Commitment
Sustained, deep research effort across the full period rather than a weekend sprint.
Agency
Independent problem-scoping and the ability to overcome research roadblocks.
Sharpness
Analytical rigour and self-correcting logic within the essay.
Expertise
Methodological depth and domain accuracy in the output.
Strategic Clarity
High-leverage, impactful research directions rather than restated risk arguments.

Online	Public
3 non-cash prizes	37 participants