Most compliance work starts too late. Teams build the risk file, draft the policy, assign the owner — and only then ask whether the AI practice they're governing is prohibited outright under Article 5 of the EU AI Act. If it is, none of that work mattered. The prohibited-practices gate has to come first, before everything else.
ProhibScan is a tool I built for that one gate. You describe the AI practice and it screens it against Article 5's prohibited categories. Nothing more. It's decision-support, which means it helps you think, not a stamp that says "cleared."
Why a gate, not a suite
Article 5 is a list of things banned in the EU: certain social scoring, manipulative systems, untargeted facial-image scraping, emotion recognition in the workplace, and similar. The job is binary in shape — prohibited or not — but the edges are subtle, and the cost of guessing wrong is high. A tool that does only this, and says so, is more useful than a broad platform that buries the prohibition question three menus deep.
The output is meant to be read by a person who then makes the call. If the practice looks close to a prohibited category, the tool points at the category and the relevant concern rather than rendering a comfortable "you're fine."
What it deliberately does not do
- No legal advice. It screens against Article 5's text; whether your specific setup crosses the line is a decision for a qualified human.
- No full Act coverage. It is the prohibited-practices gate only. Risk classification, transparency, literacy — those are other tools, other articles.
- No monitoring or certification. One pass on what you describe, not a standing clearance.
- No guarantee of completeness. If your description is vague, expect a vague read with the ambiguity called out, not a confident verdict.
ProhibScan runs in the browser for $9/mo. If it flagged something you thought was safe — or waved through something you feared — that tension is the conversation worth having before you build.
Top comments (0)