How to meet it
OWASP AISVS v1.01
Training Data Integrity & Traceability · Training Data Origin & Data Security
Protect training data in storage and transit
Verify that data integrity is provided when training data is stored and transferred.
OWASP AISVS v1.01, requirement 1.1.3
In plain English
Protect training data from being altered while it is stored or moved, and be able to prove it has not changed since it was approved.
What good looks like
- Datasets are stored with checksums or signatures recorded when they are approved.
- Transfers use authenticated, encrypted channels.
- The training job verifies checksums before it reads the data.
How teams usually do it
- Content-addressed storage or a manifest of SHA-256 hashes, checked at the start of each run.
- Object storage with versioning and write-once retention for approved datasets.
Evidence an auditor would accept
- Dataset manifests with hashes.
- Training logs showing the hash check passed.
Common mistakes
- Hashing the data but never checking the hash before training.
- Approving a dataset in one bucket and training from a copy in another with no link between them.
Not sure which requirements apply to your system? Take the assessment.
Requirement text from the OWASP AI Security Verification Standard, licensed under CC BY-SA 4.0. This guidance is shared under the same licence.