Prohibited practices — and the borderline cases companies walk into
A short list of AI uses is banned outright. Almost nobody plans one of them; plenty of teams build towards one by accident.
Last reviewed:
Sound familiar?
- An engagement feature quietly exploits a vulnerability of a specific user group.
- A workplace tool infers emotional state from voice or video 'for quality purposes'.
- A scoring feature aggregates unrelated behaviour into a general trustworthiness score.
- A vendor demo shows a capability nobody has checked against the ban list.
What we do
Screening against the ban list
We take your live systems, your roadmap and your vendor features and check them against the prohibited categories.
Borderline findings
Where a feature is close to the line, we describe what would push it over and what would keep it clearly inside.
Redesign options
Concrete alternatives that keep the business value without the banned mechanism — usually a scoping change, not a rebuild.
Documented decision
A written record of what was checked and cleared, so the next product review does not start from zero.
Questions we get
Is emotion recognition always prohibited?
No — the ban targets specific contexts, notably workplace and education settings, with narrow exceptions. That is exactly why context, not the technology label, decides the answer.
We bought the feature. Are we still liable?
Deploying a prohibited practice is not excused by having purchased it. Procurement questions belong before the contract, not after the launch.
How fast can you screen our roadmap?
A first pass over a product portfolio usually takes a few days. Anything flagged gets a deeper look with the team that owns the feature.
Tell us what's running in production.
We'll tell you what we'd check first — and what we wouldn't bother with.
Book a call