Lab

Building reliable AI for European compliance

Beyond the general‑purpose model

SinergIA Lab is where we share what we are building for the SinergIA platform. We are working on two projects: a benchmark that measures how reliable a model is on governance, risk and compliance, and a model specialised in exactly that.

General-purpose models write well and cite badly. A bank needs answers that point to the right article, run on European infrastructure and send data nowhere. That is what we work on.

In progress

The GRC benchmark

Real compliance questions on DORA, NIS2, AML, the AI Act and ESG, each with the correct answer and article. We use it to measure every model before it goes into SinergIA, and to see where it goes wrong.

In development

A model specialised in GRC

We specialise open models on regulatory texts: EU regulations, Banca d'Italia circulars, EBA and ESMA guidelines. The result can run on-premise or on European cloud, and we only adopt it when it beats general-purpose models on the benchmark.

Talk to the team

Stay connected

Previews of our research before it reaches SinergIA.