NIST NCCoE Genomic Data PETs Testbed & Dioptra Webinar

NIST's PETs and AI security testing infrastructure signals maturing tooling for privacy-preserving AI — relevant context for APS agencies handling sensitive datasets.

Key points

• NIST NCCoE is showcasing its PETs Testbed and Dioptra AI security testing platform in a June 2026 webinar.
• The testbed evaluates differential privacy and federated learning for protecting sensitive data during AI model training.
• Primarily a US-focused event; limited direct APS applicability unless agencies work with sensitive data AI pipelines.

Summary

NIST's National Cybersecurity Center of Excellence (NCCoE) is hosting a June 2026 webinar covering its Privacy-Enhancing Technologies (PETs) Testbed and the Dioptra software testing platform for AI model security evaluation. The testbed focuses on federated learning and differential privacy applied to genomic data, including results from a 2025 red-teaming exercise. Dioptra is designed to facilitate attack-and-defence evaluations for machine learning systems. The session is primarily aimed at soliciting feedback on future use cases rather than publishing finalised guidance.

Implications for Australian agencies

• Monitor Agencies developing AI pipelines over sensitive datasets — health, biometric, or otherwise — may want to monitor NIST's PETs Testbed outputs and Dioptra tooling as evidence of maturing privacy-preserving AI infrastructure.

Implications are AI-generated. Starting points, not advice.