# What is Right To Erasure?

Also called Right To Be Forgotten, Deletion Right.

The right to erasure is a data protection right allowing an individual to request deletion of personal data an organization holds about them. Established in Article 17 of the GDPR and echoed in other regimes, it is qualified rather than absolute: several grounds must apply, and exemptions exist for legal obligations, defense of legal claims and certain public interest purposes.

Erasure is triggered by specific conditions, such as the data no longer being necessary for its original purpose, consent being withdrawn where consent was the basis, the person objecting with no overriding legitimate ground, or the data having been processed unlawfully. An organization must assess the request rather than delete on demand, and must respond within statutory deadlines even when refusing.

The engineering difficulty is that personal data spreads. A single record may exist in a primary database, a replica, a backup snapshot, an analytics warehouse, a search index, a vector store, an email service and a set of application logs. A credible erasure process needs an inventory of these locations and a defensible policy for each, including how backups are handled when immediate surgical deletion is impractical.

AI systems add a harder question. Once personal data has influenced model weights, deletion of the source record does not remove that influence. Machine unlearning research aims to address this, but current methods are approximate and their guarantees are actively debated. In practice most providers avoid the problem by excluding customer data from training or by holding personal data in retrievable stores that can be deleted directly.

Erasure must also propagate to recipients. Where data was disclosed to processors or downstream services, those parties generally need to be informed of the request so they can act. Documented deletion, with evidence of what was removed and when, is what turns a stated policy into something an auditor or regulator can accept.

## Key points

- Erasure is qualified, with defined grounds and exemptions
- Requests must be answered within statutory deadlines, even if refused
- Data spreads across replicas, backups, indexes and logs
- Removing training influence from model weights remains unsolved
- Deletion must propagate to processors and downstream services

## In practice

A former user asks a company to delete their account data. The team removes the account record, purges the associated documents from object storage, deletes the derived vector embeddings, redacts the user's messages from retained transcripts, and issues a deletion instruction to the email provider that held the address. Encrypted backups still contain the record until they age out, so the response explains that retention window and the date it ends.

## Related terms

- [GDPR](/en/glossary/gdpr)
- [Data Subject Access Request](/en/glossary/data-subject-access-request)
- [CCPA](/en/glossary/ccpa)
- [Data Minimization](/en/glossary/data-minimization)
- [Audit Trail](/en/glossary/audit-trail)

[Back to the AI Glossary](/en/glossary)