Skip to content

Demo site. Fictional data; nothing is sent to anyone. Open the guided demo

UltoPulto
Try the scan

AI requests

Requests to the companies that train and run AI models, takedowns from the datasets they learn from, and signals against future training. It does not, and cannot, change a model that is already trained.

What it covers

AI model providers
Eight providers in the current registry. Each gets an access, deletion and objection request under your law, by email or as prepared portal steps.
Datasets
Takedown requests to dataset hosts for entries that point to your images.
Future training
Requests to be added to providers' suppression lists. Guided do-not-train registration is paused: the registry it used (Spawning) was offline when we checked on 4 October 2026.

How it works

  1. One request per providerOpened when your plan starts. No finding is needed to exercise the right.
  2. What the request asksWhat they hold, deletion from datasets and indexes they control, objection to training, and suppression going forward.
  3. Portal-only providersSome accept web forms only. You get the steps and the exact text to paste; the case tracks it.
  4. Deadlines and repliesReplies are logged against the case. Silence past the deadline triggers a follow-up, then a complaint draft.

What is built

Each capability carries its status. What the labels mean

  • Requests to AI model providersBuiltAccess, deletion and objection requests to AI companies, by email or as guided portal steps with the letter prepared.
  • Dataset takedown requestsBuiltTakedown requests to dataset hosts. Guided do-not-train registration is paused: the registry it used (Spawning) was offline when we checked on 4 October 2026.
  • AI dataset index matchingIn rolloutCompares a face signature with an index built from public dataset URL lists. The index is being built and reviewed.
  • Follow-ups and regulator complaint draftsBuiltA follow-up after the legal deadline, then a complaint draft for the right regulator for you to review and sign.
  • Image cloakingPlannedManaged protective changes to images before you post them.
  • Content credentials with a do-not-train signalPlannedNot built.

Limits

What this does not do.

  • No service can remove a person from the weights of an already-trained model.
  • Providers rarely disclose training data image by image, so a specific model's use of a specific photo usually cannot be confirmed.
  • Do-not-train signals depend on the trainer honouring them.
Try the scan