AI is changing how organizations use data, but it's also creating new risks along the way. Sensitive data flows into model pipelines, training sets, third-party APIs, and generative outputs—potentially leaking private information in large language models and making companies face regulatory scrutiny around AI decision-making. Traditional security frameworks weren't built for this.
Experts joined DBTA’s webinar, Data Security and Governance for the AI Era, to discuss clear, actionable strategies for protecting sensitive data, ensuring responsible AI usage, and maintaining compliance.
It is a business imperative to develop AI features into enterprise apps while ensuring compliance and security, stressed Steve Karam, principal product manager, AI, SaaS, and Growth, Delphix by Perforce.
Sensitive data discovery and masking is the gateway to compliant, predictive, and cognitive intelligence. Models need realistic but fictious data to learn, Karam said. Data versioning is required for full ai stack alignment to ensure data provenance is assured for audit, analysis, and testing.
According to Sami Akbay, VP, product management, insightsoftware, the Logi Data Platform is AI-native and future-ready, offering natural language and generative insights; open and modular; and governed and trusted.
“Analysts explore. Developers extend. Product teams embed. Business users trust. Logi products unite them all with full flexibility,” Akbay said.
Devon Kerr, director of threat research, Elastic Security, Elastic, laid out several steps to starting secure and staying secure. These steps include the following:
- Determine suitability: RAG isn’t necessary for everyone, don’t train on your sensitive data without first knowing how you plan to benefit from it.
- Segment populations: Create and assign user groups so that you can apply controls around data segmentation.
- Segment data If you do one thing, make sure you don’t allow all users to access all data. That’s how you get ants and how you wind up with data privacy issues.
- Instrument: There are low- and no-cost tools for monitoring your AI solutions, including security-specific logic to identify threat activity in this new data source.
For the full webinar, featuring a more in-depth discussion, Q&A, and more, you can view an archived version of the webinar here.