Part 3: AI Data Security & Data Security Posture Management (PR.DS)
Data Is the Foundation of AI
Artificial intelligence systems depend heavily on data. Training datasets, model outputs, and user prompts may contain sensitive business information.
The Protect function of NIST CSF 2.0 emphasizes safeguarding data from unauthorized access and misuse.
Protecting AI Training Data
Organizations should implement controls that protect data used to train AI models, including:
Encryption of sensitive datasets
Access control policies for training environments
Data classification and labeling
Monitoring of data pipelines
Compromised training data can lead to model poisoning, where attackers manipulate AI behavior.
Data Security Posture Management (DSPM)
As AI systems grow more complex, many organizations are adopting DSPM platforms to gain visibility into sensitive data across environments.
DSPM capabilities help organizations:
Discover sensitive data across cloud and AI platforms
Monitor how data is used within AI systems
Identify unauthorized data exposure risks
Enforce governance policies
Securing the data that powers AI is essential to maintaining trustworthy and reliable AI systems.
William Tulaba is a cybersecurity executive and security engineering leader focused on enterprise security strategy, cloud risk, and security operations.