Politica AI · reclamepenet

This policy explains how reclamepenet.ro uses artificial intelligence to generate content, what transparency obligations we observe, and what rights and responsibilities you have as a user. Please read it together with the Terms & Conditions and the Acceptable Use Policy.

The Romanian version at /politica-ai is the canonical text. This English translation is provided for convenience; in case of discrepancy, the Romanian text governs.

1. In short

reclamepenet uses artificial intelligence to generate advertising scripts, synthetic voices, and avatar animations. The results are labelled as AI-generated through invisible C2PA metadata embedded in the final MP4 file. At the moment of publishing to the public, we ask you to enable the platform-specific visible label (TikTok, Meta, YouTube) — see /ghid/disclosure-ai.

2. What the AI does

reclamepenet uses AI models used internally as proprietary tools for the following actions:

Generating advertising scripts (title, hook, body, CTA)
Synthesising the narrative voice (synthetic voices labelled generically)
Animating the video avatar (proprietary licensed avatars, not real persons)

The full names of the sub-processors used for these actions — together with the country of processing, the purpose, and the data processing agreements (DPAs) — are published exclusively on the dedicated page at /confidentialitate/sub-procesori, in line with the layered-notice mechanism recommended by the EDPB.

3. Mandatory labelling (AI Act Art. 50)

In accordance with Article 50 of EU Regulation 2024/1689 (AI Act), every MP4 file generated through reclamepenet carries an invisible C2PA manifest (Coalition for Content Provenance and Authenticity) with at least the following fields:

generator = reclamepenet.ro
isAIGenerated = true
generation date and time (UTC);
internal variant identifier (for audit).

This manifest is recognised by major platforms (TikTok, Meta, YouTube, LinkedIn) and enables automatic detection of AI provenance. For the visible labelling guide at publication — platform-specific toggles, recommended phrasing, examples — see /ghid/disclosure-ai.

Intentional removal or modification of the C2PA manifest from delivered files constitutes a breach of the Terms and may lead to account suspension.

4. Limits of AI generation

AI models can produce non-factual statements or hallucinations — text that is plausible but factually incorrect. This is an inherent limit of the technology, not a reclamepenet-specific issue.

Before publishing, the user must verify all specific-result claims in the generated script — in particular:

percentages or numbers (e.g. “lose 5 kg in 30 days”);
financial returns or profit guarantees;
medical, therapeutic, or wellness claims (“cures”, “treats”, “prevents”);
price or efficacy comparisons against competitors;
uniqueness, authority, or certification claims (“the only one”, “approved by”, “recommended by experts”).

Cross-link to /utilizare-acceptabila for the full regime governing user responsibility over published claims.

5. Rights over AI output

The user holds the commercial-use rights over videos generated through reclamepenet after payment is finalised and the variant is accepted. You may use the result commercially, distribute, modify, and include it in any advertising campaign, subject to restrictions propagated by upstream providers (see Terms, section 13.3).

reclamepenet claims no rightsover accepted output. For rejected previews (variants marked as refused by the user during the review window), the user's rights are void — non-accepted variants are deleted from our storage after 24 hours, in accordance with the retention policy in the Privacy Policy.

6. Data used for training

We do NOT use your content — scripts, voices, avatars, products, images, briefs, URLs supplied — to train our own AI models or those of our sub-processors. This commitment applies both to accepted outputs and to rejected previews, both to text inputs and to multimedia inputs.

All sub-processors listed at /confidentialitate/sub-procesori are bound by contractual clauses that exclude reuse of user content for training, in line with our GDPR policy (Art. 28 EU Regulation 2016/679 — processor obligations). Contractual changes that would permit training would be notified as a material change at least 30 days in advance and would require explicit re-acceptance.

7. Voice and avatar

The voices and avatars offered in the reclamepenet interface are labelled generically— for example “Voce Standard RO 1”, “Casual Gen-Z feminin”, “Avatar Profesional 2”. This labelling:

respects licensing agreements with upstream providers (named only at /confidentialitate/sub-procesori);
ensures the reclamepenet brand consistency;
does not represent identifiable real persons — synthetic voices are produced from voice models, and avatars are licensed as proprietary renderings, not portraits of real persons.

The BYOA (Bring Your Own Avatar) and personal voice cloning features are deferred to v1.1+ and are not currently supported. When introduced, they will come with additional consent and identity-verification flows, in accordance with AI Act Art. 50(4) (deepfake) and the GDPR.

8. Reporting problematic AI content

To report videos generated by reclamepenet that are used unlawfully — for example non-consensual deepfakes, disinformation, copyright infringement, impersonation of identifiable real persons — use /raportare, the DSA Art. 16 notice-and-action mechanism. We acknowledge receipt within a maximum of 24 hours and issue a reasoned decision within a maximum of 5 business days.

Alternatively, write directly to legal@reclamepenet.ro.