What happened
Cyberhaven launched Agentic Data Security (ADS) on Nov 5, 2026, a new product in Cyberhaven Flow extending Data Security and Insider Risk protection to AI apps and agents. It inventories agentic applications on endpoints (detecting models, free vs enterprise plans, extensions, tools, and unsanctioned apps), monitors all MCP tool calls over stdio or network, and reviews prompts, responses, tool calls, and file reads to stop sensitive-data exfiltration and insider risk — built on the same data lineage as its human DLP/DSPM/IRM.
Why it matters
Uses unified data lineage and DLP policy as the control surface for agents rather than a separate siloed AI-security product, reflecting the view that humans and agents act in tandem on endpoints. Addresses real concern about agents exfiltrating sensitive data or taking harmful actions at machine speed.
Applicability
Data-security and insider-risk teams evaluating endpoint controls for AI agents should pilot ADS and compare against agent-aware DLP approaches.