AI Transparency Notice

1. Where AI is used

The following features send text you provide to a large language model and return generated text or a score. Everything else in the product — templates, layout, PDF rendering, billing, the job tracker — is conventional software with no AI component.

• Résumé import and parsing: turns an uploaded PDF or DOCX into structured résumé fields.
• Bullet optimizer and bio summary: rewrites the experience bullets and summary you have written.
• Cover letter generation: drafts a letter from your résumé and a job description you supply.
• ATS check, résumé optimizer, and keyword scanner: compare your résumé against a job description and return a match score, matched and missing keywords, and written recommendations.
• Interview preparation: generates likely questions, STAR story prompts, company research summaries, and written feedback on your practice answers.
• Voice interview practice: a spoken, real-time practice session in which an AI asks interview questions, transcribes your spoken answer, and comments on its content.

2. Our role under the AI Act

EliteResume AI is the provider (Art. 3(3)) of the AI systems described above: we place them on the market under our own name. We are also a downstream deployer of general-purpose AI models supplied by third parties — the generation runs on large language models hosted on Amazon Web Services (AWS Bedrock); some fallback or development environments may route to another hosted model provider. We do not train, fine-tune, or otherwise develop foundation models ourselves, and we are not a provider of a general-purpose AI model under Arts. 51–55. When you use these features you are the deployer in your own private capacity, which under Art. 3(4) falls outside the deployer obligations of the Regulation.

3. Risk classification

We have assessed these systems against the AI Act's risk tiers and record the outcome here so you can check our reasoning.

• Prohibited practices (Art. 5): none. The systems do not perform subliminal manipulation, exploit vulnerabilities, social scoring, predictive criminal assessment, untargeted facial-image scraping, emotion inference in the workplace or in education, biometric categorisation by sensitive attributes, real-time remote biometric identification, or generation of sexual imagery of any person.
• High-risk, Annex I (Art. 6(1)): not applicable. The product is not a safety component of a regulated product.
• High-risk, Annex III point 4 (Art. 6(2)): not applicable by intended purpose. Annex III point 4(a) covers AI used to place targeted job advertisements, filter applications, or evaluate candidates in recruitment and selection; point 4(b) covers decisions on promotion, termination, task allocation, and monitoring of workers. Our systems are supplied to the candidate, for drafting and rehearsing their own application. They produce no output about any other person, are not sold or made available to employers, recruiters, or staffing agencies for candidate screening, and our Terms prohibit that use. Should we ever offer an employer-facing screening product, it would be a different intended purpose, would be classified as high-risk, and would carry the full Chapter III obligations.
• Limited risk (Art. 50): applicable. The systems interact directly with people and produce synthetic text, so the transparency duties in sections 4 and 5 below apply. We apply them now rather than waiting for the 2 August 2026 date on which Art. 50 becomes enforceable.
• The ATS match score is a keyword and structure comparison against a job description you paste in. It is a writing aid, not a prediction about you, and no employer receives it.

4. You are told when you are dealing with an AI (Art. 50(1))

Every AI-driven surface carries a visible marker. The voice interview practice session states before it starts that the interviewer is an AI system, not a person, and that the session is a rehearsal with no bearing on any real application. Requesting the microphone is a separate, explicit step, and audio is captured only while the session is running.

5. AI-generated content is marked (Art. 50(2))

Generated text is labelled as AI-generated where it is shown to you, before you decide to use it. Documents you export also carry a machine-readable marking in the document metadata recording that the file was produced with an AI-assisted tool. Note the limits of this marking: once you paste generated text into another document or a job portal, the marking does not travel with it, and after you have edited a draft the file is a mix of your writing and AI output. You remain free to use the content — it is yours — but you are also responsible for its accuracy.

6. Human oversight

The product is built so that a person — you — decides every outcome. No suggestion is written into your résumé automatically: each one appears in a review panel with an explicit Apply action, and every applied change can be undone. Scores and recommendations are advisory and change nothing in your documents. Exports and sharing happen only when you trigger them. If a suggestion is wrong, misleading, or offensive, discard it and tell us — see section 11.

7. Known limitations

Language models generate plausible text, not verified fact. Please read every draft before using it.

• Generated text can invent employers, dates, tools, metrics, or credentials that you never had. Using such content in an application is a misrepresentation and can cost you the role. Check every fact.
• ATS scores are estimates produced from the job description you supply. They are not scores from any specific applicant-tracking system, and no employer uses them.
• Quality varies by language. English input generally produces the strongest results.
• Interview questions and feedback are practice material, not a prediction of what any employer will ask or conclude.
• Speech transcription in voice practice can misrecognise words, particularly with background noise, accents, or technical vocabulary.
• Models are trained on historical text and can reproduce the stylistic and occupational stereotypes present in it. If a suggestion reads as biased, discard it and report it.

8. What the AI receives and what happens to it

Only the text needed to answer your request is sent to the model: the relevant résumé section, the job description you pasted, or the spoken answer you recorded. Your content is not used to train third-party models. We log AI usage as token counts and operational metadata for quota and abuse control, not the text of your prompts or the generated output. Retention, deletion, and the full list of processors are described in the Privacy Policy, which also covers your rights under the GDPR.

9. What these systems never do

• Infer or score your emotions, mood, personality, or confidence from your voice, face, or word choice (Art. 5(1)(f) and Art. 50(3)).
• Categorise you by biometric data, or infer race, political opinion, trade union membership, religion, sex life, or sexual orientation (Art. 5(1)(g)).
• Rank or score you as a person, or share any score with an employer (Art. 5(1)(c)).
• Make, or automate, any decision about hiring, promotion, or termination. No output of ours is a decision that produces legal or similarly significant effects for you.
• Scrape facial images or build any biometric database (Art. 5(1)(e)).
• Generate images of any kind, sexual or otherwise. The systems produce text and speech only.

10. AI literacy (Art. 4)

Art. 4 requires providers to take measures so that people working with their AI systems understand them well enough to use them sensibly. In practice that means this notice, the plain-language labels on every AI surface, the limitations listed in section 7, and the in-product guidance that consistently asks you to review generated text against your real experience before applying. Our team receives the same material. If any part of it is unclear, tell us — that is a defect in the notice, not in your reading of it.

11. Questions, corrections, and complaints

Write to support@eliteresume.net with the feature involved and, if you can, the text that was wrong. We investigate reports of inaccurate, biased, or unsafe output and correct the prompts or guardrails behind the feature. Under Art. 85 of the AI Act, anyone may also lodge a complaint about an AI system with the market surveillance authority of their EU Member State; that right is independent of contacting us first, and exercising it costs you nothing here.

12. Changes to this notice

This notice is versioned. We update it when a feature changes what the AI does, when a new AI-driven feature ships, or when the applicable rules change — including the phased application of the AI Act itself, whose Art. 50 transparency duties apply from 2 August 2026. The version and date at the top of this page always reflect the notice as it applies today.