From MIT Technology Review · · 1m 10s
AI agents leak secrets without a trace
As enterprises deploy AI agents in daily workflows, they cannot track where sensitive data enters systems or how it spreads across agents and files.
In brief
As enterprises deploy AI agents in daily workflows, they cannot track where sensitive data enters systems or how it spreads across agents and files. Enterprises using AI agents in workflows cannot trace how sensitive data moves between systems, requiring new data lineage approaches to secure agentic AI at scale…
A sales proposal leaks secrets
A sales representative asks an AI agent for customer pricing.The agent pulls data from restricted pricing systems, internal communications, and archived quotes.A second agent uses the output to draft a proposal.No human checks either handoff.
“no one can trace where the problematic data entered the workflow or how it spread across systems, files, and agents”
The problem is data visibility
As enterprises deploy agentic AI in daily workflows, they must manage moving and fragmented data continuously ingested, copied, summarized, transformed, reused, and redistributed in new forms.
“AI security isn't a tool problem; it's a data visibility problem”
Traditional security approaches fail
Traditional security approaches built solely around applications, systems, networks, or endpoints are no longer sufficient. Securing agentic AI at scale requires understanding how data moves through human, agentic, and hybrid workflows.
Data lineage is the answer
Teams need to establish data lineage to effectively manage governance, accountability, and trust across human, agentic, and hybrid workflows. This allows enterprises to trace how data moves and prevent sensitive information from spreading undetected.
The headline
Enterprises using AI agents in workflows cannot trace how sensitive data moves between systems, requiring new data lineage approaches to secure agentic AI at scale.
