AI and transparency
We use artificial intelligence at specific points in our business. Here is where, what it means for you, and the regulatory framework that applies.
Last updated: 30 September 2026. Translation of the French original.
Key points
- In our relationship with you, artificial intelligence is used in three identified places: the online audit, the search for public information about a company, and the assessment of job applications.
- Actyve trains no model. Through programming interfaces, we call models hosted and operated by Scaleway, a company established in France: it is our only artificial-intelligence provider.
- A person on our team makes the decisions that matter: accepting or rejecting a file, inviting someone to a meeting, offering an interview, hiring.
- The applicable framework is Regulation (EU) 2024/1689, known as the Artificial Intelligence Act. We follow its application timetable, as amended by Regulation (EU) 2026/1744.
- One of our systems, the assessment of job applications, falls into the high-risk category of the Regulation. We state this here rather than leave it out, and we explain below what it covers.
- You can ask at any time to deal with a person rather than a system, at contact@actyve.ai.
Where we use artificial intelligence
This page covers the artificial-intelligence systems that Actyve uses in its relationship with its visitors, prospects and candidates. The table below describes each of them: what the system does, what it produces, and who decides next. It reflects the state of our tools on 1 August 2026.
| System | What the system does | What it produces | Who decides next |
|---|---|---|---|
| Online audit: the interview | Conducts a written interview, question by question, taking your previous answers into account to formulate the next one. | Questions and rephrasings. No score and no judgement are produced at this stage. | You. You choose what you answer, you can rephrase, and you can stop at any time. |
| Online audit: writing the report | Analyses all your answers as well as the public information gathered about your company. | A written report (summary of your situation, avenues for work) and a fit score from 0 to 100 with a verdict on how well you match our criteria. This verdict makes booking a meeting easier, and opens it directly when it is favourable. It does not, however, close any door: if your profile is not selected, you can write to us to request a review by a member of the team. | A member of the team rereads every audit and decides whether to accept or reject it. The invitation to book a slot is sent by a person: it is never sent automatically. |
| Search for public information about the company | Based on the name and website you provide, searches the public web and reads the pages found. | A factual summary of the company’s business, used to ask relevant questions and to write the report. | No one, at this stage: this summary entails no decision by itself. It feeds the interview and the report, which are reread by a person. |
| Assessment of job applications | Analyses the answers to the application form as well as the public data of the GitHub account the candidate provides, if they provide one. | A score from 0 to 100, a level, a recommendation and a list of reasoned reservations. | A member of the team examines each application and decides whether to accept it, reject it or offer an interview. The score is displayed to them and helps them prioritise files; it is they, not the system, who make the decision. |
None of these systems falls under the practices prohibited by Article 5 of the Regulation, applicable since 2 February 2025: we carry out no social scoring, no emotion recognition, no biometric categorisation and no remote biometric identification.
You are talking to an AI system, and you must know it
Article 50 of the Regulation requires the provider of an artificial-intelligence system intended to interact directly with natural persons to ensure that those persons are informed that they are interacting with an AI system, unless this is obvious to a reasonably well-informed person. Its paragraph 5 specifies that this information must be provided in a clear and distinguishable manner, at the latest at the time of the first interaction. These obligations apply from 2 August 2026: they were not postponed by Regulation (EU) 2026/1744.
The rule we apply to the systems described on this page: when one of them conducts an exchange with you, you are informed before the first question, in the interface itself and not in a separate document. This is the case for the online audit interview, whose entry screen announces an audit conducted by artificial intelligence before any question is asked.
The application form, on the other hand, is not conducted by a model: its questions are written in advance by our team. It is the assessment that follows the submission of the form that involves an AI system, and it is described in the table above.
Would you rather talk to a person than to a system? Write to contact@actyve.ai.
Our role under the Regulation
Actyve designs the systems described on this page itself: the instructions given to the models, the scoring grids and the decision logic are written by our team. It puts them into service under its own name and its own brand. Within the meaning of the definitions in Article 3 of the Regulation, Actyve is therefore both provider (Article 3, point 3) and deployer (Article 3, point 11) of these systems. We do not present ourselves as a mere user of tools bought elsewhere: responsibility for what these systems produce lies with us.
However, Actyve does not train, fine-tune or develop any model. The models we call are developed and operated by third parties, which we address through programming interfaces, request by request. Actyve is therefore not a provider of a general-purpose AI model within the meaning of Chapter V of the Regulation.
Classification of our systems
The Regulation classifies AI systems by level of risk. Here is where ours stand, and why.
The online audit. This system assesses the commercial fit between a company and our offer, based on the answers of the person completing the audit and on public information about the company. It does not fall under any of the cases listed in Annex III of the Regulation. It does not evaluate the creditworthiness of a natural person or establish their credit score, which point 5(b) of that Annex covers. Nor does it determine access to an essential public service or to a social benefit, which point 5(a) covers. What is assessed is a company, and what is at stake is entering into a business relationship with us. This system is not high-risk.
The assessment of job applications. This system automatically assigns a score, a level and a recommendation to an application received in response to a job offer. This is exactly what point 4(a) of Annex III covers, which concerns systems intended to be used for recruitment, in particular to analyse and filter job applications and to evaluate candidates. It is therefore a high-risk system within the meaning of Article 6(2) of the Regulation.
We do not seek to rely on the exemption provided for in Article 6(3). The last subparagraph of that provision closes the door: an AI system referred to in Annex III is always considered high-risk where it performs profiling of natural persons. Scoring a candidate is profiling. We say so because that is the law, and because a page claiming otherwise would be worthless.
What this implies, and when. The obligations attached to the high-risk systems of Annex III do not yet apply. They were due to apply on 2 August 2026; Regulation (EU) 2026/1744, published in the Official Journal of the European Union on 24 July 2026 and in force since 27 July 2026, postponed them to 2 December 2027. These obligations cover in particular the system’s technical documentation, the description of the scoring logic, the traceability of assessments, information to candidates and formalised human oversight. We are preparing for that deadline.
A point that matters about the nature of this system: it eliminates no one. It produces a reading aid, added to the complete file in our recruitment interface. Every application sent to us is recorded there and can be consulted, including one ruled out by an elimination criterion, and it is a member of the team who reads it and decides. The score guides the reading order; it pronounces no rejection. Here is what is already in place, today, without waiting for the 2027 deadline:
- The only elimination criteria of the application form are deterministic computer rules, written in advance and applied without any artificial-intelligence model. They concern factual incompatibilities, for example a candidate who sets fully remote work as a firm condition when the position is on site, or who is looking for a desk job when the role involves meetings in the field. They apply before any call to a model.
- These criteria are announced to you before they are applied: the form explains the incompatibility and lets you change your answer rather than ruling you out silently.
- The decision to accept an application, reject it, offer an interview or hire belongs to a person on our team. It is taken after reading the file and recorded in our tools as data separate from the score. We do not claim that this person ignores the score: it is displayed with the file, and its very purpose is to sort and prioritise the applications received.
- The score does not by itself trigger a final rejection or an interview offer. Those actions are triggered by a person. It does, however, have a real effect that you should know about: below a threshold, the application remains recorded and can be consulted, but it is not flagged to the team.
What happens to your data when a model processes it
A single artificial-intelligence provider is involved for all our services: Scaleway, through its Generative APIs offering. Scaleway is a company established in France, in the European Union, and the models we call are hosted and operated by Scaleway on its own infrastructure. We go through no aggregator or intermediary, and we make no direct call to the creator of a model.
According to its documentation, Scaleway does not store, read or use the content of our requests to train, retrain or improve models; that content is accessible neither to the creators of the models nor to third parties. The same documentation provides for an exception: when requests cause abnormal errors or relate to malicious activity, their content may be kept for up to two weeks, solely to analyse and fix the problem. This is Scaleway’s commitment, published in its documentation, and not a guarantee we are able to verify ourselves.
On transfers: these requests are processed in France, within the European Union, and Scaleway states that the data concerned is stored in its Paris region. They are not transferred outside the European Union.
The models we call are chosen from Scaleway’s catalogue, whose list and characteristics Scaleway publishes in its documentation (the “Supported models” section of Generative APIs). Changing model does not change provider: the request is still processed by Scaleway, under the conditions described above.
The limits of our systems
- A report written by a language model may contain errors, approximations or poorly supported statements. Read it with that in mind, and tell us what seems wrong to you.
- The audit report is neither a certified audit nor legal, accounting, tax or financial advice. It entails no obligation of result on our part.
- The fit score measures the match between your situation and our offer. It is neither an assessment of your creditworthiness, nor a diagnosis of your company’s financial health, nor an estimate of its value.
- The public information gathered about your company comes from the web: it may be out of date, incomplete, or concern a company with a similar name. It is not checked item by item by a person before it feeds the interview.
- An application score reflects what the file made it possible to establish at a given moment, based solely on the information provided. It does not say what a person is worth, and it is not the decision.
Your rights
We do not claim that no automated processing takes place: that would be false. The drafting of the audit report, the fit score and the application score are produced automatically, and elimination criteria automatically rule out certain applications. This processing makes our sorting work easier; it does not have the final word. An unfavourable score does not close our team’s door to you: in all cases, you can write to us and ask a person to review your file. That is precisely why the following rights are open to you.
- Obtain human intervention: write to contact@actyve.ai, and a member of the team takes over the file.
- Request a review of an assessment: set out your point of view, provide the information the system did not have, and contest the result. We review it.
- Deal with a person rather than a system, at any stage.
- Obtain explanations of the logic applied: this page gives a general description, and we answer more specific questions on request.
Your rights of access, rectification, erasure, restriction, objection and portability regarding your personal data, as well as the remedies before the supervisory authority, are described in our privacy policy.
Our teams
Article 4 of the Regulation, applicable since 2 February 2025, requires providers and deployers of AI systems, including small businesses, to take measures to ensure, to their best extent, a sufficient level of AI literacy of their staff and other persons dealing with the operation and use of these systems on their behalf, taking into account their technical knowledge, experience, education and training and the context in which the systems are to be used.
Actyve makes sure that the people who operate these systems on its behalf know how they work, what they produce and their limits. The systems described on this page are designed and operated by our technical team, which itself writes their instructions, scoring grids and decision logic.
We do not claim a certified training programme here: we have not formalised one to date. This page will be updated when that arrangement is.
Contact
For any question about this page or about one of the systems described, or to request human intervention:
Actyve SRL, Rue Delwaide 60/5, 4681 Oupeye, Belgium. Company number BE 1015.964.340.
For details of the data we process and of your rights, see our privacy policy.