Italy deepfake offense: 1-5 years, and avatar consent records

A secondary source says Italy's Law 132/2025 entered into force on Oct 10, 2025 and that its Art. 612-quater deepfake offense carries one to five years. This is not legal advice; if you make avatar videos of real people, keep written consent for each person and store it with the avatar handle.

What the source reports

The point below is from a secondary source, not from the Gazzetta Ufficiale text. Read the law and ask Italian counsel what the offense covers.

Where avatars come from

Sume avatars can be made three ways: from a text prompt, from structured traits (props), or from a reference image (photo). The first two do not depict a real person. The image route does if the photo is of one, which is where consent matters.

Sume's avatar docs describe the routes and the handle but, in the pages reviewed, do not describe a consent workflow, so the record is yours to keep.

A consent record that travels with the avatar

Keep a short record per photo-based avatar, stored by avatar_handle: who the person is, what they agreed to, which uses and channels, the date, and how to withdraw. Add the job id of the avatar creation and of each video made from it. Job ids come back with every Sume job, so the chain from consent to published video can be reconstructed.

If consent is withdrawn, stop generating with that handle and remove the person's published videos where you control them.

• Photo-based avatars: written consent first.
• Store the record by avatar_handle.
• Log the job id of every video made.
• Stop using a handle when consent is withdrawn.

{
  "avatar_handle": "founder_intro",
  "person": "Name of the person depicted",
  "consent_document": "storage reference to the signed form",
  "permitted_uses": ["product intros", "paid social"],
  "date_signed": "2026-10-01",
  "avatar_job_id": "job_...",
  "video_job_ids": ["job_...", "job_..."],
  "withdrawn_on": null
}