AI Data Security
Joint guidance for protecting training, testing, and operational data throughout the AI system lifecycle.
Published by New Zealand National Cyber Security Centre, NSA, and international partners
How to use it
Use this reference when collecting datasets or connecting an AI application to business records. Record who owns each data source, who can change it, and what evidence shows that the data is still the expected version.
Protect training and testing data as separate assets, and set access rules for inputs used in production. Keep a record of significant changes so a team can trace unexpected model behavior back to a data or pipeline update.
- Access
- Public guidance