From the notebook
Are KYC verification photos deleted? How long identity data is kept, and a retention policy in code
What happens to your selfie and ID photos after KYC: why some records must be kept for years under anti-money-laundering rules while raw images can often be deleted much sooner, and a tested Python retention policy with legal holds and abandoned applications.
Not immediately, and not always. When you verify your identity with a bank or app, the company typically keeps two different kinds of data for different lengths of time. Records of who you are and how you were verified usually have to be kept for years: anti-money-laundering rules based on the FATF Recommendations require customer due diligence records to be kept for at least five years after the business relationship ends, and many countries set their own period. Raw images and video (the selfie, the document photos, the liveness recording) are a different question: privacy laws such as the GDPR require personal data to be kept no longer than necessary, so a well-run system deletes them once they’re no longer needed, unless a regulator requires them. Below is how a retention policy separates the two, as tested Python. The specific periods vary by country and regulator, so treat the numbers as examples, not legal advice.
What a KYC provider stores
| Data | Why it’s kept | Typical retention driver |
|---|---|---|
| Name, date of birth, document number, address | Proof of who the customer is | AML record-keeping: years after the relationship ends |
| The decision and its reasons | To show an auditor how the check was done | AML record-keeping |
| Selfie, document photos, liveness video | Evidence during review and disputes | Often shorter: privacy minimisation, unless rules require images |
| Abandoned applications | Nothing, once the person gives up | Delete soon |
Some regulators require copies of the document images themselves to be kept with the record, so the “shorter” row isn’t universal. The right answer is per country, and per product.
The policy in code
from dataclasses import dataclass from datetime import date, timedelta @dataclass class Case: case_id: str outcome: str # "approved" | "rejected" | "abandoned" decided_on: date | None # None while in progress relationship_ended_on: date | None = None # account closed; None while still a customer legal_hold: bool = False # Example policy. The real numbers come from your regulator and lawyers, per country. POLICY = { "raw_images_after_decision": timedelta(days=30), # selfies, document photos, liveness video "abandoned_after_start": timedelta(days=30), # never-finished applications "aml_records_after_relationship": timedelta(days=5 * 365), # FATF: at least five years } def retention_actions(case: Case, today: date, started_on: date) -> list[str]: """What may be deleted for this case today.""" if case.legal_hold: return [] actions = [] if case.outcome == "abandoned": if today - started_on >= POLICY["abandoned_after_start"]: actions.append("delete everything") return actions if case.decided_on and today - case.decided_on >= POLICY["raw_images_after_decision"]: actions.append("delete raw images and video") if case.relationship_ended_on and today - case.relationship_ended_on >= POLICY["aml_records_after_relationship"]: actions.append("delete identity records and decision log") elif case.outcome == "rejected" and case.decided_on and today - case.decided_on >= POLICY["aml_records_after_relationship"]: actions.append("delete identity records and decision log") return actions
What the tests confirm:
- An approved customer’s images are deleted 30 days after the decision, but their identity record is kept for as long as they’re a customer, even nine years later.
- After the account closes, identity records go five years after closure, not five years after sign-up.
- An abandoned application is deleted entirely after 30 days.
- A case under legal hold (an investigation or dispute) is never deleted until the hold is lifted.
What makes deletion actually happen
- Run it as a scheduled job, daily, with a log of what was deleted and why. “We’ll clean up later” never happens.
- Delete from every copy. Object storage, backups (by expiry), search indexes, analytics exports, and the logs where personal data leaked in. Keep the logs clean in the first place: see redacting personal data from logs and LLM prompts.
- Keep proof without the image. A hash of a deleted file still lets you detect if the same file is uploaded again, without keeping the file. The same goes for perceptual hashes of ID photos.
- Keep the decision trail tamper-evident for the years it must be kept, as with a hash-chained audit log.
If you’re the customer
You can ask the company, and usually its KYC provider, what they hold and for how long. Under the GDPR and similar laws you have a right of access and, where no legal duty to keep the data applies, a right to erasure. Expect the answer that core identity records must stay for the regulatory period; it’s reasonable to ask whether the images themselves are still needed.
Summary
- KYC data isn’t one thing: identity records and raw images have different retention reasons.
- AML rules (FATF: at least five years after the relationship ends) keep identity records; privacy rules push to delete what isn’t needed.
- Count retention from the end of the relationship, honour legal holds, and delete abandoned applications quickly.
- Automate deletion, and delete from every copy.
Related: how KYC identity verification works and KYC vs AML vs KYB.
Part of the guide KYC and identity verification engineering, which puts related articles in reading order.