Work on AI safety research

Find a bootcamp

Where our alumni are now

Recommended by 98% of participants

Free for you

Food, accommodation, and teaching are all provided at no cost to participants. Travel costs may be reimbursed if they're a barrier to attending

Close to home

Everyone in a bootcamp comes from the same region so you'll be learning alongside people you can stay in touch with

Great People

Work alongside talented and motivated people who are committed to making an impact

In English

All sessions, materials, and discussions are conducted in English

All backgrounds welcome

We look for people from all walks of life who are committed to contributing to AI safety

A focused environment

Accommodation is on-site and participants stay at the venue for the duration of the bootcamp

Curriculum

Themes

Technical Foundations

Build transformers from scratch, LLM agents, interpretability techniques, RLHF, evaluations

Strategy

AI capabilities and trends, risk modelling, alignment and control, forecasting, tradeoffs between mitigations

Technical Governance

Technical works for governance, compute governance, recent developments

Impact-focused projects

Formulation of your Theory of Change, literature review afternoon, 2-day capstone project with mentorship and peer feedback

Typical schedule

08:00

Breakfast

09:00

Lectures or discussions

Breaking apart technical concepts, technical frameworks, case studies, group debates

11:00

Coding Practicals

Notebooks with multiple difficulty levels

13:00

Lunch

14:00

Discussion session

Group discussion on AI Safety risks, capabilities and strategies

16:30

Break

17:00

Projects & Careers

Self-reflection, group projects, and 1-on-1 career mentorship

19:30

Dinner & Evening

Guest speakers, Q&A sessions, and social events

Eligibility

AI is going to transform every part of society, and getting it wrong could be catastrophic. Helping to steer it towards a safer outcome will require people of every background. Our technical AI Safety bootcamp is geared towards those with a technical background who are committed to contributing to AI safety in a substantial way, either through pivoting to full-time AI Safety work, building the AI Safety ecosystem, or otherwise contributing through their work or free time.

Some coding experience is required useful to make the most of the coding sessions during the bootcamp, which comprise about ~25-30% of the curriculum. Our Python workshops have multiple levels of difficulty, so they can be challenging and valuable to those with varying levels of experience.

We're most excited about people who are ready to contribute to AI safety, be that someone with decades of work experience, someone who has just finished their master's or PhD, or someone early in their career.

We expect participants to have basic familiarity with the major risks from AI (e.g. misuse, loss of control) and a rough overview of some proposed solutions. We provide a prerequisite reading list and notebooks to give everyone enough shared understanding to make the most out of the camp.

Upcoming bootcamps

Hong Kong - October '26

With support from AI Safety Hong Kong

Oct 17 - 25

Meet our teachers

Diego Dorn

Teacher

Diego is a Senior Software Developer at the PEReN, working with the EU AI Office to build the technical infrastructure required for the large-scale evaluation of models. After participating in the very first ML4Good, Diego has taught at 8+ bootcamps and is now training the next generation of teaching staff. He holds a Master's from EPFL and completed his thesis on LLM agent monitoring at CeSIA.

Julian Schulz

Teacher

Julian is a Visiting Researcher at Meridian in Cambridge, leading a project on encoded reasoning. He participated in ARENA and MATS, doing research on Steering Vectors, and worked as an independent researcher on automated feature labeling and robustness of sleeper agent detection.

Rich Barton-Cooper

Teacher; Research Manager

Rich is a Research Manager at MATS Research. He worked as a software engineer before transitioning into AI safety by completing AI Safety Fundamentals in 2024 and ML4Good in 2025, and worked on black-box scheming monitoring in MATS for 6 months before joining full-time as a Research Manager.

Charbel-Raphael Segerie

Co-founder, Curriculum Developer

Charbel is the Executive Director of CeSIA. He organized the Turing Seminar (MVA Master's AI safety course), initiated the ML4Good bootcamps, served as TA for ARENA and MLAB, and previously worked as CTO of Omnisciences and researcher at Inria Parietal and Neurospin.

Past bootcamps

Germany - May '26

South Africa - April '26

France - April '26

Singapore - September '25

UK - September '25

Germany - September '25

Brasil - September '25

Italy - May '25

Colombia - April '25

France - March '25

Germany - October '24

UK - September '24

Brasil - July '24

France - June '24

UK - March '24

Switzerland - September '23

Germany - August '23

France - August '23

Switzerland - February '23

France - November '22

Interested in bootcamps in your region?

No bootcamp in your area yet? Tell us where you are. We prioritise new locations based on interest from potential participants.