Article 5 of the EU AI Act is the part you don't want to be on the wrong side of. It lists the AI practices banned in the EU — social scoring, untargeted facial-image scraping, manipulation that bypasses human judgement, and so on. It's short, stark, and easy to read past if your product doesn't feel like it's doing anything dramatic.
The danger isn't the obvious cases. It's the feature that quietly crosses a line because nobody mapped it to the banned list. I built ProhibScan to make that mapping a five-minute check instead of a panicked one.
30-second takeaway
- Article 5 is a fixed list of bans, not a vibe — which means it's checkable.
- The risk is the feature that "doesn't feel like" a banned practice but maps to one.
- A pass today doesn't mean a pass forever; the list and your product both move.
"It doesn't feel illegal" is not a control
Most prohibited-practice violations I've seen weren't reckless. They were a growth experiment, a clever onboarding trick, a scoring widget — each innocent in isolation, each landing squarely in a banned category once you read the actual text. The teams building them weren't reckless; they were unread.
That's the whole problem with compliance work that lives in a PDF nobody opens. The law is precise enough to check against, and that's exactly why it should be a checklist you run, not a document you signed.
What a scan actually gives you
Not a legal opinion — a structured read of your system against each banned practice, with the ones you're clearly clear of, the ones that need a human look, and the ones that should scare you. Pair it with the broader AI Act and DPIA tooling; Article 5 is the first gate, not the only one.
ProhibScan runs in the browser, no account, no key. Walk through your AI system and it tells you where you sit relative to the bans.
FAQ
What is ProhibScan?
A scanner that maps your AI system against Article 5 of the EU AI Act — the list of prohibited practices — and flags where you might be exposed. No account, no key.
Does a clean scan mean I'm compliant?
It means you're clear on Article 5 specifically. The Act has more gates (risk classification, DPIAs, transparency). Treat this as the first check, not the last.
Should legal still review?
Yes. This is a triage signal that tells you where to point a human, not a replacement for one.
Author's Process
This product was built with AI coding tools. The worthwhile part was encoding the banned-practice list as checkable criteria instead of prose nobody re-reads.
Top comments (0)