Teradata has announced new privacy, security, and resiliency capabilities aimed at enterprise AI workloads. The company argues that securing data at the boundary is no longer enough when sensitive information moves across users, applications, models, and agents. Its approach combines reduced identity exposure, controls enforced closer to the data, and protections that keep AI-relevant data available and recoverable.
On the privacy side, Teradata is integrating Anonomatic's poly-anonymization technology, which replaces PII fields with stable identifiers while storing the underlying identity separately in a post-quantum-secure vault. Authorized users can restore identity when needed, but others can still work with data that retains its analytical structure and relationships. Teradata also offers format-preserving encryption for values whose format matters to downstream systems. The company says large-scale testing showed protected multi-table joins used fewer resources than equivalent unprotected queries, though results vary by workload.
For security and resiliency, Teradata adds post-quantum protection for data in transit via TLS 1.3 hybrid-key exchange, along with dynamic data masking, classification and tagging, column-level encryption, and global identity. A partnership with Immuta brings attribute-based access control that applies policy consistently across users and AI workloads. Immutable backups prevent recovery points from being modified or deleted, and database-level high availability adds an extra layer for mission-critical workloads. Teradata's chief product officer, Sumeet Arora, said privacy, security, and resiliency have historically been treated as separate problems, but that approach does not hold up as AI environments grow more complex. Anonomatic CEO Matt Fleck added that enterprise AI forces organizations to use sensitive individual-level data, and the goal is to avoid choosing between protection and utility.