AI-generated content policy

Transparency about AI use

[Configurar en admin] expressly declares, in compliance with Article 50.2 of the European AI Act, that the site publishes text and image content generated with the assistance of artificial intelligence. Each entry in the catalog is the result of editorial prompts, assisted generation and final human curation. This document describes the scope, the models used, the review measures and the correction mechanisms.

Which content is AI-assisted

The following elements of the site are generated or assisted with AI:

• The descriptions, lore, attributes and descriptive fields of each entry are drafted with the assistance of language models, starting from documentary sources and entity-specific prompts.
• The representative images of each being are generated through diffusion models, based on prompts describing documented mythological traits (not biographical traits of real people).
• Synonyms, etymologies, suggested taxonomies and tentative relationships are proposed with AI assistance and reviewed by humans before publication.
• Translations between the 10 supported languages are performed with the assistance of multilingual models and reviewed by humans in critical languages for the target audience.

Model providers

The site uses, at different stages of the editorial flow, the following artificial intelligence providers:

• models from the Grok family (xAI) for assisted drafting and curated translation tasks.
• image-generation models via cloud provider, configured with safety filters and internal mythological prompt policies.

Human curation before publication

No entry in the catalog is published without first going through human editorial review. [Configurar en admin] applies the following minimum review protocol:

• verification of the consistency of the entry with the declared cultural sources (comparative mythology, etymology, taxonomy).
• detection of cultural biases, stereotypes or inappropriate representations, with special attention to non-Western traditions.
• manual expansion or correction of the cited bibliographic sources.
• rejection and regeneration when the result does not meet quality, documentary accuracy or cultural-sensitivity criteria.

Accidental resemblances with real people

The images published in represent exclusively mythological, folkloric or demonological entities. The prompts used start from documented cultural sources (archetypal traits, iconographic attributes, classical literary descriptions) and do not include any reference to real persons, living or deceased. There is no intent to portray, satirize, impersonate or link to any identifiable person.

Any resemblance between the generated images and real people is mere statistical coincidence inherent to the operation of generative models.

Diffusion models compose images from patterns learned over millions of human images, which can produce, in an unsought and uncontrollable manner, facial features reminiscent of a specific person. [Configurar en admin] does not use prompts referring to real people, does not associate the results with personal names and keeps the original prompts as documentary evidence of the exclusively mythological intent.

Good-faith takedown procedure

If you recognize yourself or a third party in an image generated by the site and consider it to affect your image rights, honor or reputation, [Configurar en admin] applies the following good-faith takedown procedure, without any admission of responsibility or intent:

1. Send a justified notification to [email protected] indicating the URL of the entry, a description of the perceived resemblance and, where applicable, reasonable evidence of your identity or representation.
2. The editorial team confirms receipt within a maximum of 72 working hours and internally reviews the original prompt, the declared cultural sources and the documentary context of the image.
3. If an identifiable problematic resemblance is confirmed, the image is removed from the site within a maximum of 7 calendar days from the notification.
4. A new image is regenerated for the affected mythological entity, adjusting the prompt parameters to avoid the resemblance, and republished as the corresponding illustration.

We keep the original prompt and editorial flow as documentary evidence that the intent was exclusively mythological and not biographical. To speed up the process write to [email protected].

Ownership of the generated images

Images generated on commission for [Configurar en admin] with its own editorial prompts are published as illustrations for the catalog. Ownership and reuse are subject to the terms of the corresponding AI providers and to the applicable legislation on assisted-generated works. For third-party uses, see our editorial licensing policy on the press page.

Errors, biases and corrections

We know that AI can introduce historical errors, cultural biases or inaccurate descriptions despite human review. If you detect a documentary error, an etymological inaccuracy or a culturally inappropriate representation, write to [email protected].

Use of site content for AI training

[Configurar en admin] does not authorize the use of the content published on (curated texts, images, taxonomies, relationships, translations) to train third-party generative artificial intelligence models. This restriction is additionally published through machine-readable metadata and is part of the site's terms of use.

Changes to this policy

This policy will be updated as the providers used, the human review measures and the applicable regulation evolve. The date of last revision appears at the top of the document.