---
title: "Output suppression"
description: "A filter around a model that stops it producing a person's data: the fast first answer to an erasure or objection request when the data sits in the weights and retraining is disproportionate."
canonical: https://aigovernanceengineer.com/glossary/output-suppression
author: "Jorge García Aibar"
license: "CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/)"
doi: https://doi.org/10.5281/zenodo.22956197
version: "0.5.0"
updated: 2026-09-25
---

# Output suppression

A filter around a model that stops it producing a person's data: the fast first answer to an erasure or objection request when the data sits in the weights and retraining is disproportionate. The CNIL accepts filters shown to be effective and robust and prefers general rules to a list of names [1]. The data stays in the model.

- Developed in: [ch. 19, Suppression, retraining and unlearning](https://aigovernanceengineer.com/bok/privacy-and-ai#suppression-retraining-and-unlearning); [ch. 05, Pattern: Rights Requests Against Models](https://aigovernanceengineer.com/patterns/rights-requests-against-models)
- Chapters: [ch. 05, Patterns](https://aigovernanceengineer.com/bok/patterns) · [ch. 19, Privacy & AI](https://aigovernanceengineer.com/bok/privacy-and-ai)
- Contrast with: [Machine unlearning](https://aigovernanceengineer.com/glossary/machine-unlearning)
- In the glossary chapter: https://aigovernanceengineer.com/bok/glossary#t-output-suppression

## Sources

[1] "Ensuring and facilitating the exercise of data subjects' rights" (AI how-to sheet; retraining; output filters accepted if shown sufficiently effective and robust, based on general rules rather than lists of people). CNIL. 2026-01-05. https://www.cnil.fr/en/respect-and-facilitate-exercise-data-subjects-rights (verified: primary)