We have no customers yet, so this page starts with what does not need one. Everything in the first four sections is on this site already and can be checked without asking us anything.
Twelve questions a reviewer would ask, answered in public
A reviewer wrote down twelve questions cold. The answers are published as given, and the answer to four of them is no. Five carry a dated correction that keeps the wording it replaced, so you can see what we got wrong and when we fixed it.
We ran the engine on ourselves and published what it got wrong
We answered our own engine's questions about our own product, ran the classification and published the output verbatim — including the sentence it got wrong about us. A vendor's engine run on the vendor is the only test where the vendor cannot choose the input.
The legal texts we derive from, pinned by sha256
Every classification is derived from source files published here with their sha256 sums. If one of those files differs from what we pinned, our build fails. You can download them and check the hash yourself.
Every assessment records which version of the law it rested on
The Act has already been amended once. An assessment that cannot say which text it answered is a filing cabinet, not evidence. Each of ours carries the corpus it was derived against, and the amending act is published here point by point.
Complipath is EU AI Act compliance software for AI-heavy software companies without a compliance team — an AI system register, deterministic risk classification, the obligations that follow, and the evidence behind every decision.
Complipath is built by Yobel Tzegai in Gothenburg, Sweden.
Complipath provides legal information, not legal advice. Every guide cites its source on EUR-Lex — Regulation (EU) 2024/1689, and Regulation (EU) 2026/1744 where that has amended it; where the law is still settling, the guide says so.
We measure page views with Vercel Web Analytics. It uses no third-party cookies. Visitors are identified by a hash derived from the incoming request, which is discarded after 24 hours, and no identifier is stored that could follow a visitor to another site. What is collected: the time of the visit, the URL, the referring page, filtered query parameters, city-level location, operating system, browser and device type.