Compliance

Our legal stance

AJUTIA builds and offers AI tools. Under the AI Regulation, this may involve provider and deployer obligations; who holds each role is not fixed in advance and depends on the system, the configuration and the contract. The formal determination is pending legal advice. This page explains what we do and what we do not do.

Provider vs deployer

Provider and deployer are two roles foreseen by the AI Regulation, but who holds them cannot be fixed in advance: it depends on the system, the configuration and the contract. AJUTIA may act as provider in some cases and as deployer in others; the same applies to the client. The distribution of duties is assessed for each specific use; the formal determination of who is responsible for each obligation is pending legal advice.

Prohibited practices (art. 5)

Article 5 of the AI Regulation prohibits certain practices in the cases it defines: inferring emotions in the workplace or at school, except for uses for medical or safety reasons (art. 5(1)(f)); categorising people from biometric data by sensitive attributes (art. 5(1)(g)); and social scoring that leads to detrimental treatment that is unjustified or disproportionate (art. 5(1)(c)). These cases apply since February 2025. We do not offer them. Verifying they are not used depends on the deployer's configuration and use; each specific case must be examined.

AI notice (art. 50.1)

The obligation to inform that you are talking to a machine. It must be visible and at the beginning, not hidden in the terms of use. It is a design and QA criterion that every AJUTIA assistant shows the notice in the first message; transversal deployed verification is pending.

High risk (annex III)

Systems with reinforced requirements: personnel selection, education, access to public services, creditworthiness assessment. According to the European Union calendar, these requirements apply from 2 December 2027. We are working to get ahead of them.

Human oversight

A person approves before anything irreversible happens: sending to the administration, charging, publishing. AI proposes; the person decides. You can configure that no irreversible action happens without human approval. You need to enable and use it that way.

We are not auditors

We do not certify anything. We do not promise you will pass any audit. The goal is to help you arrive with ordered evidence, not to guarantee the outcome. The difference matters: we give you the tools; the auditor is who verifies.

Cycle

Continuous review and improvement

We accompany the review of obligations and improvement with human decisions:

1Observe: gather evidence of what each tool does and how it is used.
2Interpret with evidence: compare the observation with applicable obligations.
3Propose options: present alternatives without imposing a decision.
4Decide: the responsible person chooses what to do; the tool does not decide for them.
5Apply with authorization: execute the decision with the corresponding consent and permissions.
6Check and review: verify the result and start again if needed.