New capabilities help security teams prevent sensitive data from being ingested into AI memory, then continuously govern exposure across copilots and agentic AI.
SAN FRANCISCO, Calif., March 23, 2026 /PRNewswire-PRWeb/ -- Lightbeam, the identity-centric data security and privacy platform powered by the Data Identity Graph, today announced new AI security capabilities to help enterprises secure AI agents across Microsoft Copilot, ChatGPT Enterprise, and Google Gemini.
As enterprises move from AI assistants to agentic AI that can take autonomous actions across systems, the risk model becomes more complex. Security teams are no longer protecting only data at rest and in motion. They also need to protect data as it gets pulled into AI contexts, summarized, recombined, and retained in AI memory. That requires two controls working together: AI information governance to classify and label sensitive data before it becomes AI-accessible, and AI usage control to continuously monitor AI interactions, enforce policy, and stop unsafe behavior as AI agents operate at machine speed.
Gartner has highlighted both the governance gap and the threat trajectory for agentic AI. "For many organizations, weak and fragmented information governance is emerging as the major obstacle to wider GenAI rollouts." Gartner also warns that "Through 2029, over 50% of successful cybersecurity attacks against AI agents will exploit access control issues, using direct or indirect prompt injection as an attack vector."
Lightbeam turns these AI security capabilities into operational controls that help teams regain control and governance over AI usage in their organization.
AI information governance that starts before the first prompt
Lightbeam discovers and classifies sensitive data across structured and unstructured repositories, maps it to identities and data subjects, and applies enforceable labels that help organizations prevent sensitive data from being used in AI retrieval, agent memory, and other AI workflows where it does not belong.
This release includes Smart Classify, a new AI-security-focused capability that accelerates "AI-ready" classification and labeling. Smart Classify learns patterns in high-risk enterprise content that agents commonly touch, such as tax information, customer records, financial reports, contracts, and recurring internal templates, and then applies consistent classification and "No AI" style labeling at scale. The result is fewer blind spots, less manual review, and more reliable enforcement when AI agents attempt to retrieve or retain sensitive information.
AI usage control guardrails for Agentic AI
Lightbeam continuously monitors AI interactions and inspects prompts and responses for policy violations, abuse patterns, and attempts to extract sensitive data. When high-risk exposure occurs, Lightbeam routes alerts into governance workflows and drives remediation, including least-privilege access changes and, where applicable, suspending or revoking access to sensitive information that should not be reachable through AI. This gives security leaders practical control over oversharing and misuse, even as AI becomes embedded in everyday workflows.
Expanded coverage across enterprise GenAI tools
With this release, organizations can apply a consistent label-first governance and usage control approach across Microsoft Copilot, ChatGPT Enterprise, and Google Gemini. Customers can standardize how sensitive data is excluded from AI workflows, and how AI interactions are monitored and governed, even as teams adopt multiple copilots and agents across the business.
Availability
AI information governance and AI usage control are available for Microsoft Copilot immediately. Expanded support for ChatGPT Enterprise and Google Gemini will be available in April 2026.
Resources
- Lightbeam AI security solutions page
- Invisible Technologies written case study
- Invisible Technologies video highlights:
- Blog: AI Data Exposure Starts Before the First Prompt
About Lightbeam
Lightbeam is an identity-centric data security and privacy platform that helps organizations prevent breaches, reduce ransomware risk, and automate compliance by unifying data security, governance, and privacy operations. Powered by the patented Data Identity Graph, Lightbeam discovers and classifies sensitive data across structured and unstructured environments and maps it to real identities and business context to enable precise controls, continuous monitoring, and automated remediation.
Media Contact
Seth Knox, Lightbeam, 1 650-224-8706, [email protected], Lightbeam.ai
SOURCE Lightbeam
Share this article