Fair use

Our fair-use approach

Localign sells per-seat licences. Fair use means: one seat, one human; reasonable, professional volumes per person. This page explains what we consider fair, what we don't, and how the product behaves when an account is approaching its monthly limit.

Trust center

What fair use means here

Each seat is for one named person inside your organisation. Heavy daily use by that one person is welcome — that is what the licence is for. What fair use prevents is sharing a seat across multiple humans, or using a single seat to drive industrial-scale automation that would normally require an enterprise pricing tier.

What is fine

Heavy individual usage across the working day. Batch summarisation of your own documents and emails. Multi-language workflows where one user moves between Dutch and English content all day. Creating and using as many internal assistants as your organisation needs. Using the product on a desktop and on mobile from the same seat. None of this triggers fair-use action.

What is not

Sharing a single login across multiple humans. Scripted bulk querying that bypasses the seat model — for example, automating thousands of requests per day from a single seat to feed an external system. Reselling Localign output as a standalone service. Using Localign output to train a competing AI model. Any of the practices prohibited under article 5 of the EU AI Act, which the AI Annex restates explicitly.

What happens if a limit is hit

We use soft caps, not surprise lockouts. At roughly 80 percent of the monthly token allowance for your licence type, the user sees a friendly warning in the app and an organisation administrator gets the same notice in the admin panel. At 100 percent, the seat is rate-limited until the next monthly reset, with a clear path to upgrade or buy additional seats. Existing in-flight conversations are not interrupted.

Prohibited uses

On top of fair use, certain uses are prohibited regardless of volume. The AI Act article 5 list is the contractual baseline: no subliminal or manipulative techniques, no exploitation of vulnerabilities, no social scoring, no untargeted face scraping, no emotion recognition in workplaces or education, no biometric categorisation by sensitive traits, and no real-time remote biometric identification outside the article 5 exceptions. The Compliance — EU AI Act page restates the full list.