AI Transparency Statement

Jaicob B.V.

Version: 3.0, October 2026

1. About this statement

Jaicob B.V. ("Jaicob", "we", "our", or "us") operates an AI-powered recruitment platform. Artificial intelligence is at the core of what we build, and we believe that everyone who comes into contact with our platform has the right to know when AI is involved, what it does, what it does not do, and who remains responsible.

This statement explains, in plain language:

which parts of our platform use AI;
how we make sure you always know when you are interacting with AI;
how humans stay in control of decisions that affect candidates;
what our AI systems are never allowed to do; and
what rights you have and how to exercise them.

This statement is written for three audiences: candidates whose applications are handled through our platform, the companies that use our platform (our clients), and anyone else who interacts with our AI features. It supplements, and should be read together with, our Privacy Policy and Terms and Conditions.

This statement describes how our platform is designed and operated as of the version date above. It is provided for transparency purposes and does not constitute legal advice.

2. Our role under the EU AI Act

The European Union has adopted the AI Act (Regulation (EU) 2024/1689), a law that regulates how AI systems may be developed and used, as amended by Regulation (EU) 2026/1744. AI systems used for the recruitment and selection of people are classified as "high-risk" under Article 6(2) of the AI Act, read together with Annex III, point 4(a). This is because recruitment decisions can have a significant impact on people's lives and opportunities. The duties for high-risk systems take effect on 2 December 2027. The duty to tell people when they are dealing with an AI system (Article 50) already applies, since 2 August 2026.

Our platform falls within this category. Under the AI Act, Jaicob acts as the provider of these AI systems: we develop them, we document them, and we are responsible for designing them so that they can be used safely, transparently, and under meaningful human oversight.

The companies that use our platform to recruit (our clients) act as deployers under the AI Act. They decide how the platform is used in their hiring process and carry their own obligations, such as informing candidates and ensuring that qualified people oversee the system's output. We provide our clients with detailed instructions for use and compliance documentation to support them in this role.

As a provider of high-risk AI systems, we maintain, among other things:

a risk management process covering the lifecycle of our AI features;
technical documentation describing how our AI systems work;
data governance procedures for the data used by our AI features;
automatic logging of AI system activity (see Section 8);
human oversight mechanisms built into the product (see Section 5); and
a quality management system for the development and maintenance of our AI features.

3. Where our platform uses AI

Not everything in our platform is AI. Many features are ordinary software: an applicant tracking system, a CRM, scheduling, and messaging. The following features do use AI:

AI Sourcing: searching public and licensed professional data sources to identify potential candidates that match a vacancy.
AI Matching and screening: comparing a candidate's profile or application against the requirements of a vacancy and producing a match score, screening summary, or ranking as decision support for recruiters.
AI pre-selection of sourced profiles: for people found through AI Sourcing, an AI pre-selection decides which profiles are shown to the recruiter, and a match score below a fixed level keeps a workflow-sourced profile out of view. Every rejection is recorded with its reason. Workflow automations (called "Automations" in the product) then send messages written by the recruiter as templates; they contain no AI-generated text.
AI agents: a recruiting organisation can set up an AI agent (called "Studio" in the product) to find and contact candidates, answer them, ask screening questions, mark a candidate qualified, move an application to a later stage, and close an application that has gone quiet (this is not a rejection, and the organisation can reopen it). These steps happen automatically by default. For some of them, such as finding candidates, a screening conversation by text or WhatsApp, booking an intake or a client interview, replying to a client, or finding prospects, the organisation can set the agent to ask a person first instead. Other steps, such as recording a screening answer or moving an application forward, always happen automatically. Rejecting an application, disqualifying a candidate, and making an offer always wait for a person, whatever the organisation chooses (see Section 5). An AI agent's first message to you on each channel says that AI helped write it (see Section 4).
AI interviews and conversations: structured screening conversations with candidates by chat, SMS, or voice call, in which an AI system asks questions and records answers.
Jaicob: an AI assistant for recruiters that answers questions about their own data and performs tasks at the recruiter's request.
Document understanding: extracting structured information from documents such as CVs and resumes, so that candidates and recruiters do not have to re-type it.

Most AI outputs in our platform are decision support: they inform recruiters, who decide what happens next. Three features act without a person reviewing each step first: the AI pre-selection above, AI interviews, and AI agents. None of them can reject or disqualify a candidate on their own. Section 5 explains how human decisions stay in place.

4. You will always know when you are interacting with AI

The AI Act (Article 50) requires that people are informed when they interact with an AI system. We build this disclosure into the product itself, so that it does not depend on manual steps:

Screening conversations by SMS and WhatsApp: the first message you receive from the AI tells you that you are chatting with an AI assistant and that you can ask to speak to a person at any time. This line is inserted by the platform itself, not by an instruction to the model, and cannot be switched off.
Voice calls: before an AI phone interview connects you to the AI, the platform plays a fixed spoken message, naming the company, that tells you that you are talking to an AI assistant and that you can ask to speak to a person at any time. This is inserted by the platform itself and cannot be switched off. It plays in the company's language, which may differ from yours. The AI is also instructed to introduce itself as an AI assistant once the call connects.
Consent first: no AI screening conversation starts, on any channel, unless you opted in through an application form that names the AI conversation and its channels. You can stop an SMS or WhatsApp conversation at any time by replying STOP.
Messages from AI agents: when an AI agent writes to you for a recruiting organisation, its first message to you on each channel (LinkedIn, email or WhatsApp) includes a line, added by our platform, saying that AI helped write it. The organisation cannot switch this off. The line is left out when a recruiter wrote the message as a template, or approved it before it was sent.
Replies to a company's clients: when an AI agent replies, on a recruiting organisation's behalf, to someone at one of its clients, the first AI reply in that conversation carries the same line. A new conversation gets the line again. The organisation cannot switch this off.
Outreach messages: messages sent by workflow automations are written by the recruiting organisation as templates and contain no AI-generated text, so no AI system is speaking to you there. If your profile was found through a public or licensed source, the recruiting organisation decides whether its first message to you includes a notice about this, and how it is worded. Our default notice tells you who the organisation is, that your profile came from an external source, why they are contacting you, where to read more, and how to object with one click. A removal link you have received keeps working, whatever the organisation decides later.

These disclosure rules are not region-dependent. We apply them to all conversations on our platform, regardless of where you are located, because we consider them good practice everywhere (and because similar rules exist outside the EU, such as the California "B.O.T." disclosure law).

You are never required to complete a conversation with an AI system. If you prefer to speak with a human, you can say so at any point, or contact the company that reached out to you directly.

5. Humans remain in control

The AI Act (Article 14) requires that high-risk AI systems are designed so that humans can effectively oversee them. The GDPR (Article 22) additionally protects people against decisions based solely on automated processing that significantly affect them. Our platform is built around both principles:

Reject, disqualify, or hire: always a person's decision. Match scores, screening results, and rankings are advisory. Where a recruiting organisation uses an AI agent, it can contact you, answer you, ask screening questions, mark you qualified, and move your application forward on its own. Deciding to reject you, disqualify you, hire you, or make you an offer is always made by a person, not by the platform.
Rejection is a human act. The only path in our software that notifies a candidate of a rejection, sends the outcome to a connected recruitment system, or closes a candidate's journey requires a logged-in recruiter with permission, including when an AI agent proposed it. When an agent proposes a rejection, the recruiter approving it sees the exact email you will receive and can change it before it is sent. No score, filter, or unapproved agent or assistant action can reach this path.
Approval on assistant actions. When the recruiter assistant wants to message a candidate, send an email, delete a record, or commit to a multi-step plan, it asks the recruiter first, one action at a time. Approval requests are cryptographically signed, so an approval cannot be fabricated.
Readable reasons, not only scores. Every match score comes with six dimension scores and a written reason for each, and every re-score is kept as a new version, so a recruiter can read why before deciding.
Interruptibility. Clients can pause or stop automations, and candidates can end AI conversations at any time. Workflow automations act when they run; the recruiter's oversight there is in reviewing what an automation does before starting it.

We also warn our clients explicitly against "automation bias": the tendency to accept an AI suggestion without genuine review. Approving an AI proposal is a real decision, and our documentation and product design treat it as one.

6. What our AI systems never do

Some uses of AI are prohibited outright by the AI Act (Article 5), and others are excluded by our own design choices. Our platform does not, and is not allowed to:

Infer emotions. We do not use emotion recognition or emotion inference in screening, scoring, interviews, or any other recruitment feature. Emotion inference in the workplace context is prohibited by Article 5(1)(f) of the AI Act, and no such analysis exists anywhere in our platform.
Use protected personal characteristics in AI evaluation. Attributes such as gender, nationality, and date of birth are deliberately withheld from the information our AI evaluation features receive. Our AI features are also instructed not to infer such characteristics.
Perform biometric categorisation or identification. Voice conversations are processed for transcription and conversation purposes only. We do not create voice prints, do not identify people by their voice, and do not derive personal characteristics from biometric data.
Score people socially. We do not evaluate people based on social behaviour or personal characteristics unrelated to the specific vacancy.
Pretend to be human. As described in Section 4, every AI conversation or message tells you, the first time, that AI is involved. Our AI never pretends to be a person.
Make final decisions about people on its own. As described in Section 5, an AI agent can move your application forward, but deciding to reject, disqualify, hire you, or make you an offer is always made by a person.

7. Data quality, fairness, and bias monitoring

AI systems can reproduce biases present in data. We take this risk seriously, because in recruitment it translates directly into unfair treatment of real people. Our measures include:

Excluding sensitive attributes from evaluation, as described in Section 6, so that AI assessments are based on job-relevant information such as skills, experience, and qualifications.
Structured, vacancy-specific evaluation. Match and screening features assess candidates against the requirements of the specific vacancy, not against generic profiles.
Keeping the pre-selection blind to your name. The AI pre-selection of sourced profiles does not receive the candidate's name; it judges role, skills, location, and industry signals only.
Measuring for bias. Where a recruiting organisation uses AI agents and has switched on this comparison, we compare scoring outcomes across candidate groups, by sex and age group, every day, and the organisation can see the results and export anonymised scoring records in its own account. The comparison is off until the organisation switches it on, and switching it on is its instruction to us to work out sex and age group from the data it already holds, only to measure bias. Coverage is limited today, because most people found through sourcing have no gender or date of birth on record. We also run a nightly test that scores matched pairs of sample CVs differing in one detail unrelated to the job, such as a school or an association, to check whether it shifts a score in a way that could stand in for a protected characteristic. That test found small such shifts. We corrected them by no longer giving the AI those details. A small difference remains for some school names, and we check it every night.
Every decision is reconstructable. Each AI assessment is stored with its scores, its written reasons, and the exact prompt version and model that produced it, so any individual decision can be examined after the fact.

8. Logging and record keeping

The AI Act requires high-risk AI systems to automatically record events while they operate (Article 12) and requires providers to keep these logs for at least six months (Article 19). Our platform does this by design:

every match score is recorded with its dimension scores, written reasons, prompt version, and model, per version; every AI pre-selection rejection is recorded with its reason;
workflow executions, message deliveries, and an append-only audit log of access, consent, erasure, and changes to candidate data are recorded;
logs of high-risk functionality are retained for at least six months, enforced in our software as a floor that no configuration can lower; our software can remove message content once that period has passed, and we are switching this on step by step; and
logs are available to the client for inspection and to supervisory authorities where the law requires it.

Retention of logs is balanced against data protection law: where logs contain personal data, they fall under the retention and deletion rules described in our Privacy Policy, and records kept longer for audit purposes are anonymised.

9. The AI models we use

Our platform is built on large language models and other machine-learning models from established third-party providers (so-called general-purpose AI models), combined with our own logic, guardrails, and evaluation layers. About these models:

we contract with model providers under data processing agreements, as required by Article 28 GDPR;
candidate and client data submitted to these models through our platform is not used by the model providers to train their models, based on the contractual commitments in place;
we host and process data in the European Union wherever the required capability is available in the EU, and apply the international transfer safeguards described in our Privacy Policy where it is not; and
we select, configure, and test the models we use, and we remain responsible for how they behave within our platform. An up-to-date list of our subprocessors, including AI model providers, is available to clients as part of our data processing agreement.

10. Your rights as a candidate

If your application or profile was handled using our AI features, you have the following rights:

The right to know. The company you applied to (or that contacted you) must inform you that AI is used in its recruitment process. Our platform's built-in disclosures (Section 4) support this, and our client documentation requires it.
The right to an explanation. Under Article 86 of the AI Act and Article 22(3) GDPR, you can request a clear and meaningful explanation of the role AI played in a decision that affected you. Our platform keeps, for every AI assessment, the scores and the written reason behind each of them, which the recruiting company uses to answer you.
The right to human involvement. You can ask for a decision to be reviewed by a human being. As described in Section 5, a rejection on our platform is always a human act; on request, the recruiting company must also be able to explain and reconsider a decision with human involvement.
The right to opt out of AI conversations. You may decline an AI interview or AI voice call and ask for a human alternative. Declining an AI conversation may affect timing, but the recruiting company remains responsible for treating your application fairly.
Your data protection rights. Access, rectification, erasure, restriction, portability, and objection, as described in our Privacy Policy.

Who to contact. The company you applied to, or that contacted you, decides how your data is used and is your first point of contact for these rights (it is the "data controller" under the GDPR and the "deployer" under the AI Act). If you contact us instead, we will help route your request and provide the recruiting company with the technical information needed to answer it. We will never leave a candidate's rights request unanswered because of this division of roles.

Complaints. You can lodge a complaint about the use of your personal data with the Dutch Data Protection Authority (Autoriteit Persoonsgegevens, www.autoriteitpersoonsgegevens.nl) or the supervisory authority in your own country. Complaints about the use of an AI system as such can be lodged with the market surveillance authority designated in your country under the AI Act.

11. What we expect from our clients

Compliance in recruitment AI is shared work. As deployers, our clients must use the platform in line with our instructions for use. In particular, clients must:

inform candidates that AI is used in their recruitment process, before it is used on them;
inform their own employees and employee representatives where required (Article 26(7) AI Act);
assign oversight of the platform to people with the competence, training, and authority to exercise it;
genuinely review AI proposals instead of approving them automatically;
use the AI features only for recruitment and selection, their intended purpose; and
keep logs and cooperate with candidates' rights requests as described above.

We publish detailed compliance documentation for clients, including instructions for use and a deployer checklist, in our client documentation portal at developers.jaicob.ai.

12. Security

AI features run on the same secured infrastructure as the rest of our platform: data encrypted at rest and in transit, strict access controls, EU-based hosting, and independent security certifications (including ISO 27001:2022). The full description of our security measures, breach procedures, and certifications is in our Privacy Policy.

13. Outside the European Union

Our platform is also used outside the EU, and AI rules differ per country and per state. We host our platform in the European Union, and the AI that reads candidate data processes it there too (Section 9); where we use another provider, it does not receive candidate records, or the transfer safeguards described in our Privacy Policy apply. The other safeguards apply everywhere too, such as AI disclosure and a person on every rejection. One setting differs by region: a recruiting organisation using AI agents chooses, for itself, whether its contact data follows EU retention rules or US retention rules. Under the US choice, contact data is not deleted automatically; this does not change AI disclosure or who decides a rejection. Our AI disclosure practices are consistent with the California B.O.T. Act. Recruiting organisations using AI agents can export their scoring records, without candidate names, for their own bias review; whether this is enough for New York City Local Law 144 or a similar duty depends on the organisation's auditor and on how much data is covered. Clients remain responsible for verifying the rules that apply to their own hiring locations, and our documentation helps them do so.

14. Changes to this statement

We update this statement when our AI features change in a way that matters for transparency, when the law changes, or when we improve our practices. The version number and date at the top of this page always reflect the current version. Significant changes are announced to our clients; the AI Act's documentation requirements also oblige us to keep our instructions for use up to date.

15. Contact

Questions about this statement, or about how our platform uses AI, are welcome:

Jaicob B.V.
Burgemeester Stekelenburgplein 199, 5041 SC Tilburg, the Netherlands
Email: support@jaicob.ai
Chamber of Commerce (KvK): 92883761