AI Potluck
Contribute
Back to Gap Map Product / UX / Assurance & compliance evidence

COMPL-AI

COMPL-AI
open source / Overall score: 1.6

COMPL-AI is a compliance-centered evaluation framework for large language models: a technical interpretation of the EU AI Act paired with an open benchmarking suite, now 29 benchmarks, that covers the Act's technical requirements. It is built on the Inspect evaluation framework and publishes results on a public leaderboard. ETH Zurich, INSAIT and LatticeFlow AI maintain it.

Placed here rather than in evaluation code because its output is a report mapped to the EU AI Act. Tagged self-attested.

Openness

5 medium confidence
5.0
license
Apache-2.0(OSI)
source
public
core features withheld
no

The framework and benchmark suite are Apache-2.0 and maintained jointly by two research institutes and a company, with the whole suite in the public repository.

Adoption

1 low confidence
1.0

GitHub stars are the only signal, since the repository is the declared way to run it. A star is not a use.

Capability

2 medium confidence
2.0

COMPL-AI reports benchmark results against named EU AI Act requirements, but in its own format and run by whoever evaluates the model, so a regulator takes the numbers on trust unless it reruns them. Compliance Trestle writes to a published standard that outside tools can validate.

Verified 2026-09-26