Artificial intelligence

In one sentence

An artificial intelligence model reads the text of your CV to draw a site’s data from it. It does not assess you, does not score you, does not rank you, and nothing it produces is published until you have reread it and decided to.

What the model does

It receives the text extracted from your CV and returns a structure: your jobs, your studies, your skills, your projects, as they appear in your document.

It is put to one task, extraction, and used once, when your site is created. It plays no part afterwards: what you correct in your space, you write yourself.

What the model does not do

• it is designed and instructed to add nothing: it is told not to fill the gaps in your CV and to add no job, degree or skill you did not write. It can nonetheless get things wrong or misread, which is why nothing is published before you have reread it;
• it leaves aside what a public portfolio has no business carrying: it is told not to extract your date of birth, age, nationality, family situation, phone number or postal address;
• it does not qualify you: no judgement, no level, no summary of your worth is produced;
• it does not score you, rank you, compare you to other people or select you;
• it decides nothing: no decision producing legal effects concerning you is taken automatically;
• it publishes nothing: publishing is always an explicit click of yours.

That is also what places cvestia under the European regulation on artificial intelligence: a system that assessed, filtered or selected people for recruitment would be high risk, whose obligations apply from 2 December 2027 (Regulation (EU) 2026/1744). cvestia does none of that, and will not start without saying so here.

Nor is cvestia a conversational system or a generator of synthetic content meant for the public: the regulation’s transparency obligations, in application since 2 August 2026, do not apply to cvestia; this page says more than they require.

What is sent, and what is not

The text extracted from your CV is passed once only, when the site is created, to OpenAI Ireland Ltd and its affiliates, including OpenAI, LLC in the United States. Under its terms, this provider may keep the request for up to 30 days solely to detect abuse, then erases it. It is never used to train its models. The PDF itself is never passed on.

What never goes with it: your email address, your sessions, your site’s visit measurements, and everything you wrote or corrected after the site was created.

The API terms exclude training on what is sent to it, unless explicitly opted in, which cvestia does not do.

No training on your content

cvestia neither asks for nor grants any right to use your CV, your content or your corrections to train, fine-tune or evaluate a model, ours or anybody else’s. This is not a policy that could quietly change: it would be a right the terms of use would have to ask you for, and they do not.

The business terms of every provider that could be configured rule out training on what is sent to them, and cvestia enables no model-improvement option at any of them. The detail, provider by provider, is on the Privacy page.

At Google, that holds only of its paid services: the service refuses to start an extraction on a Google project that is not one.

The provider can change

The name on this page is read from the configuration in service: it is therefore the one in operation at the moment you read it.

The name of the provider that actually read your CV is recorded with your site and is what counts: you can ask us for it, even months later. A change of provider is announced on this page and, if you have an account, by email.

A model gets things wrong, and rereading is yours to do

A model can leave out a job, attach a date to the wrong entry, confuse an employer with a client. What the service hands you is a draft, visible to you alone, and your space exists to correct it.

You alone are responsible for what your site shows once published. cvestia verifies nothing you declare: not your jobs, not your degrees, not your certifications.

If the extraction missed something, adding it by hand from your space is usually simplest: nothing obliges you to deposit another CV.

Your site is readable by machines

A published site exposes a file readable by automated systems, and a public page can be read, analysed or copied by a search engine as much as by an artificial intelligence system. We can neither prevent it nor know who did it.

That is the price of a public address, and it is why publishing is a deliberate gesture. Until you publish, your site is visible to you alone.