Artificial intelligence transparency policy

1. Editorial responsibility

The person responsible for AI transparency and editorial control is Jean-François Callen. Questions and correction requests may be sent to contact@callen-jf.com.

2. Covered content

AI may be used to prepare drafts, suggest translations, structure text, generate illustrations or edit images. Gazette articles may inform the public about cultural matters of public interest.

3. Visible labels

• AI-generated or AI-edited illustrations carry an explicit disclosure.
• Deepfakes are clearly and recognisably labelled.
• AI-generated public-interest text without human editorial review is identified as such.
• When an AI-generated draft has been reviewed, corrected and approved by a responsible person, the review date and reviewer are recorded and may be displayed as an enhanced-transparency measure.

4. Human editorial review

Publishing AI-assisted or AI-generated text requires confirmation of human review, unless it is deliberately published as unreviewed, in which case a stronger disclosure is shown. The person approving the content remains responsible for sources, corrections and the final publication decision.

5. Machine-readable information

Each published article has a JSON-LD provenance record containing the declared origin of text and images, tools and models where known, human-review status, approval date, IPTC digital source type and a SHA-256 hash of the illustration when available.

6. Metadata, C2PA and original files

Uploaded images are stored without recompression or re-encoding, so existing metadata and Content Credentials/C2PA are not intentionally removed. The website does not claim to create a cryptographic C2PA signature; such a signature must come from the image generator or compatible signing software using a trusted identity.

7. Accuracy and corrections

Provenance information relies on declarations by authors, administrators or creation tools. Incorrect or uncertain information may be corrected, supplemented or removed. The register records key declaration and review operations.

8. Member content

Members must disclose realistic artificially generated or manipulated content that could mislead the public. Deceptive material, impersonation, unlabelled deepfakes and false information may be hidden or removed.

9. Scope

This policy is an editorial and technical transparency measure. It does not replace individual legal advice on copyright, image rights, defamation, data protection or rules applying to AI providers and platforms.