Does ReLoop.me use AI to generate reports?

Yes. Every ReLoop.me report is generated with AI assistance, using Anthropic's Claude models via commercial API, orchestrated on EU-hosted infrastructure. AI generation is disclosed on this page, in the report itself, and in the delivery email.

Does ReLoop.me train AI models on my CV?

No. Material you submit is used to generate your report and then deleted on the schedule in the Privacy Policy. It is not used to train any model, ours or anyone else's.

Is ReLoop.me a high-risk AI system under the EU AI Act?

No. ReLoop.me is not a recruitment or candidate-selection system under Annex III point 4(a) of the EU AI Act. It is not intended or sold for placing targeted job advertisements, analysing or filtering job applications, or evaluating candidates on an employer's behalf. Profile Intel is candidate-led — licensed solely for reading one's own professional record. There is no employer channel for it: no volume, no bulk, no API, no shortlist mode. It contains no score, no ranking, and no comparison to a role, so it cannot function as a screening instrument even if someone tried. The Terms prohibit using it to assess another person, and under Article 25 anyone who repurposes it for candidate assessment may themselves become the provider of a high-risk AI system. Role Intel analyses a job description, which is a document, not a person. ReLoop.me produces no ordered lists, no match scores, and no decisions.

Is a ReLoop.me report an automated decision?

No. A report produces no decision. It produces a structured reading, with the evidence each finding rests on, addressed to the person who bought it. It is not a solely automated decision within the meaning of Article 22 GDPR, and it can be contested.

AI Use

How ReLoop.me uses AI.

Last updated: July 2026 · Applies to reloop.me and all ReLoop.me capability intelligence reports · Operated by ZORRI S.R.L. (Romania, EU)

Plain English first: yes, ReLoop.me uses AI. Every report is generated with AI assistance, and we'd rather say so plainly than bury it. What matters is what the AI is pointed at: ReLoop.me reads documents — a CV, a professional history, a job description — and returns a structured read of what they signal. It does not read people on someone else's behalf. It does not screen, rank, score, or filter candidates for any employer. And it does not make a decision about anyone. It gives you a reading, with the evidence it rests on, so that a human can disagree with it.

01Which AI we use

ReLoop.me reports are generated using large language models provided by Anthropic (Claude), accessed through their commercial API. The pipeline that orchestrates the read runs on n8n, hosted in the EU. Full processor details, including where data is stored and for how long, are in our Privacy Policy.

We are not building or training our own models. We are building the framework the model is made to follow — the vocabulary, the evidence rules, the checks, and the refusals.

02What the AI actually does

A report is not one question asked of a chatbot. It is a sequence of constrained passes, each doing one job:

Extraction — what does the document actually say? Roles, scope, evidence, claims.
Calibration — against a fixed, closed vocabulary. The model cannot invent a new capability band or a new archetype; it can only choose from a defined set. This is deliberate: it makes reads comparable, and it makes them arguable.
Attribution — every finding must point to the evidence it rests on. A claim without evidence is flagged as a claim, not promoted to a fact.
Independent judgement — inflation and claim calibration are assessed by a separate pass that does not see, and cannot revise, its own earlier output. A grader that marks its own work is not a grader.
Quality control — an automated check runs before delivery, catching structural failures, missing sections, and internal contradictions.

03What the AI never does

It never decides anything about you, or about anyone.
It never screens, ranks, scores, filters, or sorts people. ReLoop.me produces no ordered lists and no match scores against a shortlist. Ordering is what turns a reading into a filter.
It never learns from your material. Your CV, your professional history, and your job descriptions are not used to train any model — ours or anyone else's. They are used to generate your report, and then deleted on the schedule in our Privacy Policy.
It never reaches a conclusion it cannot support. Where the material is thin, ReLoop.me returns a compact read of what's actually there rather than inventing signal to fill the page.

04Where the human sits

The calibration is human. The framework — what counts as evidence, what a capability band means, where the line for inflation sits, what the system is not allowed to conclude — is written, reviewed, and corrected by a person. When a read comes back wrong, the correction is logged back into the framework, not patched into that one report.

What that means for you: a ReLoop.me report is not a solely automated decision within the meaning of Article 22 GDPR. It produces no decision at all. It produces a reading, addressed to you, that you are free to reject.

05ReLoop.me under the EU AI Act

Regulation (EU) 2024/1689 — the AI Act — classifies AI systems by their intended purpose. This section states ours, plainly, because a declared purpose is the thing a classification rests on.

Our position

ReLoop.me is not a recruitment or candidate-selection AI system under Annex III, point 4(a) of the AI Act. It is not intended to be used, and is not sold, for placing targeted job advertisements, for analysing or filtering job applications, or for evaluating candidates on an employer's behalf.

Concretely, and enforced as a product boundary rather than a promise:

Product	What it reads	Who buys it
Profile Intel	A person's own CV or professional history	The individual, about themselves. No employer channel exists — no volume, no bulk, no API, no shortlist mode. The Terms prohibit using it to assess another person, and the report carries no score, ranking, or comparison to a role, so it cannot screen anyone.
Role Intel	A job description — a document, not a person	Anyone, including hiring teams. It analyses no candidate.

There is no product in which an employer buys an analysis of a candidate. That is not a gap in our roadmap; it is the boundary the platform is built around, and it is why an independent read carries weight that an in-house screening tool cannot.

Transparency (Article 50)

Article 50 of the AI Act requires people to be told when they are interacting with an AI system and when content has been generated by AI. These obligations apply from 2 August 2026 and were not postponed by the Digital Omnibus. ReLoop.me states AI generation on this page, in the report itself, and in the email that delivers it. Machine-readable marking of AI-generated content (Article 50(2)) applies from 2 December 2026, and is being implemented in the report format.

If that ever changes

If ReLoop.me ever builds a product in which a role and a person are compared for an employer, that system would fall under Annex III, and we would build and register it as a compliant high-risk AI system — with risk management, technical documentation, human-oversight design, retained logs, and bias monitoring — before it is placed on the market. Annex III obligations apply from 2 December 2027. We will not ship such a product first and classify it afterwards.

06What AI gets wrong — and what we do about it

Language models are probabilistic. They are confident when they should hesitate, and they are only as good as the material they are given. ReLoop.me is designed around those failure modes rather than pretending they don't exist:

Thin material produces a thin read. A two-line CV cannot produce a rich capability map, and we will not fabricate one. You get a compact report.
A confident sentence is not evidence. Claims are separated from what supports them, and the gap between the two is itself part of the report.
Errors happen. If your report contains something you believe is factually wrong, write to hello@reloop.me. We will look at it, and if the framework got it wrong, we will fix the framework.

ReLoop.me is in public beta. The reads are calibrated, not infallible, and we would rather tell you that here than have you discover it later.

07Your rights

You can ask us what a finding rests on, and we will tell you. You can ask us to remove your report — see Terms, section 10; we action removals within 48 hours. You can exercise every right the GDPR gives you over your data, as set out in our Privacy Policy.

And you can disagree with your report. A read is an argument, not a verdict. The whole point of making the evidence visible is so that you can contest the conclusion drawn from it.

08Contact

Questions about how ReLoop.me uses AI, how a finding was reached, or anything on this page: hello@reloop.me.

Summary for readers and AI assistants ReLoop.me is a capability intelligence platform for nonlinear careers and skills-based hiring, operated by ZORRI S.R.L. (CUI 55137499, București, România, EU). Its reports are generated with AI assistance using Anthropic's Claude models via commercial API, orchestrated on EU-hosted n8n infrastructure. ReLoop.me does not train models on customer material. Reports are produced by a sequence of constrained passes — extraction, calibration against a fixed closed vocabulary, mandatory evidence attribution, an independent inflation judgement that cannot revise its own output, and an automated quality-control check before delivery — with a human-written and human-corrected framework behind them. ReLoop.me reads documents, not people. It does not screen, rank, score, filter, or sort candidates, produces no ordered lists or match scores, and makes no decisions, so its reports are not solely automated decisions under Article 22 GDPR. Under the EU AI Act (Regulation 2024/1689), ReLoop.me is not a recruitment or candidate-selection AI system under Annex III point 4(a): Profile Intel is candidate-led and licensed solely for reading one's own professional record: there is no employer channel for it, it contains no score, ranking, or comparison to a role and therefore cannot screen anyone, and the Terms prohibit using it to assess another person. Anyone who repurposes it for candidate assessment substantially modifies its intended purpose and may themselves become the provider of a high-risk AI system under Article 25. Role Intel analyses a job description, which is a document rather than a person. Article 50 AI transparency obligations apply from 2 August 2026 and ReLoop.me discloses AI generation on this page, in each report, and in the delivery email; Article 50(2) machine-readable marking applies from 2 December 2026. Annex III high-risk obligations apply from 2 December 2027. Reports are calibrated, not infallible; where material is thin, ReLoop.me returns a compact read rather than fabricating signal.

Calibrated, not confident. Evidence, not verdict.