JobMentis and AI: what our models see, where they run, and what they never do

Which models we use, and where they run

Every AI feature goes through one gateway, so the list below is complete.

• Text: Gemini models from Google, run on Vertex AI in the European Union. Under Google Cloud's terms, your prompts and the answers are not used to train Google's models.
• If Gemini is unavailable: Claude models from Anthropic, in the United States. Under Anthropic's commercial terms, API data is not used to train its models. OpenAI can serve as a further fallback when configured, under the same default.
• Voice: ElevenLabs, in the United States, for the First CV interview, mock interviews, networking rehearsal and Story Bank dictation. It runs the speech recognition, the AI voice and the conversation, and keeps the transcript of each session.
• Images: the LinkedIn roast sends the screenshot you upload to Gemini image models through Google's Gemini API rather than Vertex AI, so it is not held to the EU region.
• ElevenLabs does not use recordings made since 18 September 2026 to train its models: that option is switched off for our account.

The full list of the companies that process your data, and where they are based, is in our privacy policy.

What the AI sees, feature by feature

Each feature sends a model only what it needs. In plain words:

What it never sees

A CV you ask us to audit or rewrite is read as a whole: it is the document you are working on. Everything else is narrower.

• Kept out of the profile summary our features share: your email address, your phone number, and any date of birth, age, marital status, children or nationality your CV states.
• Never sent to a model: the photo you add to your CV in our editor, your payment details and your Google Calendar. A photo printed inside a CV file you upload is read with the rest of the file.
• Never in the features you use: your coach's private notes.

What it will never do

Every instruction that reads your words carries the same rule: never infer, raise or take into account your health, disability, religion or belief, political opinion, union membership, ethnic origin, sexual orientation, sex life, pregnancy, family situation, age or nationality.

• If your CV mentions one of them, it is treated as ordinary content: kept as you wrote it, never explained, never expanded, never advised on unless you ask.
• None of it is ever a reason for a score, a rank, a gap or a piece of advice.
• Nothing we store labels you with one of them. The check that points them out in the CV editor runs in your browser and saves nothing.

Where the numbers come from

• Salary benchmarks on our salary pages and in Worth are computed from public statistics and fixed rules, never written by a model, and a model's text around them is checked so it cannot add a figure of its own.
• Two figures are a model's estimate: the salary range suggested for a job you track, calibrated to those benchmarks, and the market-worth estimate in Worth.
• Scores: a model rates each part (skills, experience, evidence and direction for the fit; format, keywords and clarity for the CV), and fixed weights in our code turn those ratings into the number you see.

Files created with AI

PDF and Word files created with AI carry an invisible "AI-assisted" tag in their metadata, in line with EU transparency rules. It does not appear on the page. Plain text and Markdown exports carry no tag, because a line of text would end up inside your application.

Your controls

• Your CVs, work history, stories, identity statement and answers are all in your account, where you can correct or delete them. What you delete stops reaching our AI features.
• Career Memory explains what our features remember about you and why.
• To delete your whole account, or to receive a copy of your data, write to support@jobmentis.com. Deletion runs across every system we use.

Nothing decides for you

No employer receives a score from JobMentis, and nothing here makes a decision about you. Every document is yours to edit before it goes anywhere, and the voice and chat assistants say they are AI at the start of every conversation.

Questions