Learn to evaluate AI capabilities and safety.

Structured training in evaluation methods, hands-on practice and a mentored team project—for researchers and professionals working across AI, safety and governance.

Cohort 02 · February–June 2027

Apply Now
Abstract collage of black and white photos, geometric shapes, and digital elements on a white background.

Why this Programme?

Artificial intelligence is advancing at unprecedented speed. Yet, as frontier AI systems grow more powerful, our ability to evaluate their capabilities and risks has not kept pace.

This programme exists to change that.

  • Talent Gap: AI Safety Institutes and leading labs worldwide face a shortage of experts in evaluation.

  • Unique Approach: We combine technical depth with policy and governance, bridging a gap no other programme fills.

  • Impact Pathway: Our graduates will be equipped to join top research labs, government agencies, and industry, where their skills are urgently needed.

This is the first step toward establishing AI Evaluation & Safety as a formal academic discipline - a foundation for the first MSc in the field.

Cohort 01 in numbers

Over four months, 40 participants from 18 countries worked through 12 modules, two hands-on courses, and a team-based final project, ending with an in-person week in Valencia.

Within weeks of the last session:

  • Four of our faculty were appointed to the EU AI Act Scientific Panel.

  • 16 of 28 surveyed already reported a concrete outcome: new roles including one at NIST/CAISI, a fellowship, a PhD offer, a collaboration.

  • 12 teams were taking their capstone to NeurIPS, ICLR, or FAccT.

  • 23 alumni volunteered to mentor the next cohort.

Meet Cohort 01
40participants
673applications
45+faculty
18countries
11project teams
86NPS
9.5/10would recommend
75%exceeded expectations
60%with concrete outcomes already
93%pursuing eval/safety roles

  • "I haven't felt intellectually stimulated like this for a long time. It's incredibly challenging, but also really rewarding."

  • “The programme gave me the confidence I needed in technical evaluation. As someone who was always close to evaluations but never performed one, I felt I wasn’t ‘technical’ enough to do it. The programme gave me the space to learn, and to work with others to gain the skills and knowledge needed.”

  • “Connections and community. I was someone who had a passion for AI evaluation, but with no one to learn from or discuss it with. This programme has facilitated a complete shift in career domain for me — from educational measurement to AI evaluation, which is my new home.”

  • “It helped me manage engineers building an eval at work, and I’m also including what I learned into the GenAI product roadmap.”

  • “Meeting everyone in person in Valencia. I formed meaningful connections with many people in the program, and I feel like I am now a part of the growing evaluations community.”

  • “This programme helped shape the Trust & Safety portfolio I lead at work, and will be fundamental as we contribute that work to the federal institute.”

Abstract geometric digital artwork with circles, rectangles, and lines in shades of blue, black, and gray.

Cohort 02 - Programme Snapshot

40 participants. 5 months. One goal: to make AI accountable.

Dates: February-June 2027, including onboarding and final presentations

Delivery: Online teaching, practical work and mentored team projects

Live participation: Sessions on Tuesdays/Thursdays, European morning/afternoon, live attendance expected

Commitment: 15 hours per week

Capstone project: You’ll work on a your own AI evaluation problem with a team and a mentor. A capstone sprint is planned for spring 2027. Depending on funding, it might be in-person in Valencia, Spain

Cohort 02: 40 top global participants with fully funded scholarships

Learn more about the Programme

Who’s Involved

This programme has the support of faculty from leading universities, including the University of Cambridge, Stanford, Princeton, Beijing Normal University, Renmin University of China, William & Mary, the Technical University of Valencia, University of Oxford, Cornell University, MIT, University of Bristol, DTU and TU/e.

Cohort 01 faculty also came from key institutions, research organizations, and companies such as the EU AI Office, the UK AI Safety Institute, FAR AI, RAND, Epoch AI, Apollo Research, Redwood Research, METR, Microsoft Research, and Google DeepMind.

Four faculty members appointed to the EU AI Act Scientific Panel.

Meet Our Faculty
University of Cambridge logo with a shield featuring lions and a crown, and the text 'University of Cambridge' in black.
Stanford University logo with a red 'S' and a green tree inside, and the words 'Stanford University' in red next to it.
Princeton University logo featuring a shield with an open book and Latin inscriptions on top, and black and orange chevrons at the bottom.
Oxford.png
Cornell.png
MIT.png
The official seal of the Beijing Normal University
Logo of Renmin University of China featuring three stylized red flames inside a circular badge with Chinese characters and the date 1937.
Southern University.png
William & Mary logo with a gold crown and emblem, green text that reads 'William & Mary,' and smaller green text below that reads 'Chartered 1693'.
Logo of the Technical University of Valencia
Bristol.png
DTU.png
TU e.png
Logo of the European Artificial Intelligence Office featuring a circle of stars and the initials 'AI' in the center.
Logo of AISI Safety Institute with bold red letters 'AISI' followed by the words 'SAFETY INSTITUTE' in black outline.
Black and white silhouette of a city skyline at night with skyscrapers, a bridge, and water reflections.
Purple square logo with white curved line in top left corner, and the text 'RAND CORPORATION' in white centered below the line.
EPOCHAI logo with a stylized graphic of five parallel lines in a slanted orientation next to the word "EPOCHAI" in dark blue capital letters.
Apollo Research logo, wordmark
A blue logo featuring a stylized pine tree within a circle, with interconnected nodes forming the tree's shape. Below, the text reads "REDWOOD RESEARCH" with "REDWOOD" in bold and larger font, and "RESEARCH" underneath in a smaller, capitalized font.
Microsoft Research logo with a four-color window icon and the text 'Microsoft Research' in gray.
DeepMind logo with a spiral blue symbol and the name in blue text.
METR.png

Our Academic Directors:

“Why a new AI Evaluation discipline is needed now, and what we’re doing about it.”

Sponsored by the Valencian Graduate School and Research Network of Artificial Intelligence (ValgrAI), funded by Coefficient Giving (formerly known as Open Philanthropy), and backed administratively by the Berkeley Existential Risk Initiative (BERI).

Logo for Valgrai with the text 'Valgrai' in blue and purple, and 'Valencian Graduate School and Research Network of Artificial Intelligence' underneath, with a small molecular structure icon next to the text.
Stylized black and purple logo with a purple square outline starting at the top left corner, with the word 'BERI' in bold purple letters below the square.
Abstract collage featuring a black and white moon, digital glitch textures, text snippets, ocean scene, and geometric shapes in black, white, turquoise, and orange.

Shape the future of AI evaluation.

Join the next cohort.

Get the full picture. Build the missing expertise.

This programme gives you the panorama; the tools, frameworks, and perspectives to connect the dots between machine learning, evaluation, and governance.

You’ll leave ready to:

  • Understand how AI systems are tested and validated.

  • Speak the language of both researchers and policymakers.

  • Join a network building the standards the world will rely on.

Apply Now