Lab B · 14 seats
Red-teaming your own model
A full adversarial pass over a model you bring, with a written findings report by the end of the session.
Day 02
Track 04
Workshops
Workshops are capped, hands-on and require a laptop. Seats are released to members first, then to in-person pass holders.
Studio 2 · 20 seats
Writing an ethics impact statement
Draft a statement that survives review, using real submissions from past editions.
Day 03
Lab A · 12 seats
Evaluation beyond benchmarks
Designing task suites that measure what you actually care about.
Day 02
Workshop C · 24 seats
Open-weight deployment
Serving, quantising and monitoring open models on modest hardware.
Day 03
Studio 1 · 16 seats
Interpretability by hand
Probing activations without tooling magic, from scratch.
Day 04
Lab D · 18 seats
Data provenance in practice
Tracking, licensing and documenting training corpora.
Day 04
Programme details
Plan your session
Schedule
Red-teaming your own model
Bring a model and complete a guided adversarial pass.
Evaluation beyond benchmarks
Build task suites around real operational goals.
Open-weight deployment
Deploy and monitor models on modest hardware.
Speakers & hosts
- Dr. Ines Haddad — AI evaluation
- Tomás Ferreira — Open hardware
- Maya Chen — Responsible deployment
Location
Labs A–D, Paris conference venue. Laptop required; room assignment appears after registration.