AI usage

At Lente Binaria we use artificial intelligence systems (hereinafter, AI) to speed up parts of our design, development, communication and audiovisual production work. This page explains where we use AI, how we identify content generated or assisted by AI, and what rights you have as a client or visitor.

This statement is published voluntarily ahead of the full application of Regulation (EU) 2024/1689 of the European Parliament and of the Council, of 13 June 2024, on Artificial Intelligence (hereinafter, the AI Act), scheduled for 2 August 2026, and in particular of its Article 50 (transparency obligations).

01Where we use AI

1.1 In our internal processes

• Development assistance (Claude by Anthropic, GitHub Copilot): code generation and review, audits and refactoring. All code goes through human review before being delivered to the client or published to production.
• Editorial assistance (Claude by Anthropic): text drafts, translations and transcript summaries. All final text goes through human editing.
• Transcription and subtitling (VibeVoice ASR, our own engine, and Whisper by OpenAI): converting audio to text in meetings, interviews and videos. Transcripts are reviewed manually before use.

1.2 In client deliverables

• AI-generated images (Imagen by Google, possibly other models): when used in a client's landing page, banner, mockup or graphic material, they are identified as such in the delivery dossier and, when the format allows it, carry C2PA metadata or a visible caption (“AI-generated image”).
• Video: we do not currently use AI video generation models in client deliverables. If we add this capability in the future, this statement will be updated before the first deployment, with a complete labelling process: a visible video caption, end credits, prior written notice to the client and C2PA metadata when the format allows it.
• Synthetic voice (VibeVoice TTS): when used to narrate audiovisual pieces, it is identified in the credits as a synthetic voice, unless the client holds the rights to the original voice being cloned (with their documented express consent).
• AI-generated text: no text published on a client website goes out without human reading, editing and approval. The editorial owner is always a person at Lente Binaria.

1.3 On this website

• Some texts on this website started as AI-assisted drafts, manually edited by the team. Editorial authorship belongs to Lente Binaria.
• The photographs of the team and of real work are real images, not synthetic. No portrait of the team is synthetic: the faces you see are always real photography.
• The following illustrative images on this website have been generated with AI (Google's Gemini / Imagen models) and keep the invisible SynthID watermark that identifies them as synthetic content:
  • The home page poster (hero-poster).
  • The images for the Presence, Growth and Custom packs (pack-presencia, pack-crecimiento, pack-custom).
  • The audiovisual and web service panels (panel-audiovisual, panel-web).
  • The craft band (banda-oficio) and the night view of Seville (sevilla-nocturna).
  • The decorative knurling texture and the 404 error page image (404-bokeh).

02Labelling of synthetic content (art. 50 AI Act)

Under Article 50.4 of the AI Act, those who deploy AI systems that generate or manipulate images, audio or video constituting deepfake content, or texts published to inform the public on matters of general interest, must disclose in a clear and distinguishable manner that the content has been artificially generated or manipulated.

Our labelling policy:

• Visible labelling in the caption of images, video or audio in pieces aimed at the end audience.
• Machine-readable labelling through C2PA metadata in compatible formats, when the production chain allows it, and preservation of the SynthID watermark in images generated with Google models.
• Traceability: every deliverable includes in its dossier the list of AI-generated pieces, the model used and the date.

03Human decisions versus automated decisions

At Lente Binaria, no decision affecting the end client (price, scope, deadlines, incident resolution, content published under their brand) is made in an automated way. AI speeds up tasks, but it does not decide:

• Business proposals are written and signed by a person on the team.
• Corrections after a review are applied by a person on the team.
• After-sales support is always handled by a person.

If a specific project were to deploy a system with automated decision-making (for example, a client chatbot that classifies leads), the client and the end user would be informed before its deployment, in accordance with Article 50.1 of the AI Act.

04Systems we do not use

So that what we do not do is also clear:

• We do not use unacceptable-risk AI systems (social scoring, cognitive manipulation, real-time mass biometric identification).
• We do not use high-risk AI systems (education, employment, justice, healthcare) as defined in Annex III of the AI Act.
• We do not generate content that passes itself off as human opinion without saying so (fake reviews, synthetic testimonials).
• We do not generate synthetic portraits of real people: the team's faces are always photography.

05Models and providers we use

For the sake of transparency, the main AI model providers we work with:

• Anthropic: Claude (text, code, analysis). Headquarters: USA · privacy policy.
• Google: Gemini (text/multimodal) and Imagen (images). Headquarters: USA · privacy policy.
• OpenAI (occasionally): Whisper (transcription). Headquarters: USA · privacy policy.
• VibeVoice: our own ASR and TTS engine, developed and hosted by Lente Binaria. It does not send data to third parties.

When a project involves sending personal data or confidential information to any of these providers, it is documented in the project's data processing agreement, with its legal basis, purpose and time limit. See also our privacy policy.

06Your rights

As a client or visitor, you have the right to:

• Know whether a specific piece in a deliverable has been generated or manipulated with AI.
• Ask for a deliverable to be redone without AI assistance (this may involve a longer deadline or a higher cost, which we will communicate to you beforehand).
• Receive information about the models and providers used in your project.
• Object to specific uses in your project (for example, “I do not want synthetic voices in my video”).
• Have no decision that affects you made in an automated way without your knowledge.

07How to exercise these rights · contact

• Email: ia@lentebinaria.com (alternatively: info@lentebinaria.com)
• Phone: +34 608 408 970
• Contact form: lentebinaria.com/en/contact

Maximum response time: 20 working days.

08Reviews of this statement

This page will be reviewed, at a minimum:

• Before the date of full application of the AI Act (2 August 2026).
• Annually from then on.
• Whenever a new AI provider or model joins Lente Binaria's work stack.
• Whenever the applicable legislation changes (the Royal Decree implementing the AI Act, AESIA guidelines, EDPB decisions).

Legal basis: Regulation (EU) 2024/1689 (AI Act) · Regulation (EU) 2016/679 (GDPR) · Spanish Organic Law 3/2018 (LOPDGDD). Supervisory authority in Spain: Spanish Agency for the Supervision of Artificial Intelligence (AESIA).