🕵️Unvetted AI Agents Leaked Confidential Pricing Data Into Sales Proposal Without Human Review

Your sales agents might be leaking pricing data right now

A sales rep asks an AI agent for customer pricing. The agent pulls from restricted pricing systems, internal comms, and archived quotes, then hands off to a second agent that drafts the proposal. Nobody checks either handoff. The final document ships with unapproved, confidential, and sensitive information in it. If you're building or running agentic workflows, this is your problem. The model did its job. The gap is that no one can trace where the bad data entered or how it spread across systems, files, and agents. Traditional access controls assumed data sat still. Agents copy, summarize, transform, and redistribute it continuously, so perimeter security doesn't see the leak. The core issue is data lineage. Without it, you can't answer basic questions like which source fed which output or who approved what. Accurate-looking output can still be built on incorrect or restricted inputs, which makes spot-checking the result nearly useless. Governance has to track data movement, not just tool permissions.