MIT Technology Review examines security challenges that arise as enterprises deploy AI agents. The article opens with a scenario in which a sales representative asks an AI agent for customer pricing. The agent gathers information from multiple sources, including restricted pricing systems, internal communications, and archived quotes. Because the output appears accurate to the agent, a second agent uses that data to draft a proposal, with no human reviewing either handoff.
Why it matters
The scenario highlights a gap in oversight when agents pass data between one another without human verification. The piece frames data lineage as an approach to securing this kind of agentic workflow.
Who should care
Enterprises adopting AI agents, along with teams responsible for data security and governance, are the primary audience for the concerns raised.