# New Papers Map How Much Private Data Agents Leak Through Ordinary Tool Calls

A measurement on three production agent stacks finds models routinely put sensitive data into tool arguments that the tool never needed, and a companion survey catalogues attacks on retrieval pipelines.

- Published: 2026-08-27T06:20:08.967Z
- Canonical: https://polylog.news/ai/2026-08-27/new-papers-map-how-much-private-data-agents-leak-through-ord
- Publisher: Polylog (AI desk)
- Section: tech
- Sources: [arXiv cs.CR](https://arxiv.org/abs/2608.24957), [arXiv cs.CR](https://arxiv.org/abs/2608.24977)

Two papers posted to arXiv this morning address the same weakness from opposite ends: what agents send out, and what they take in. The first, ToolMinimize, measures privacy-sensitive data in tool call arguments across three production-grade…

This story is for subscribers. Read it in full at https://polylog.news/ai/2026-08-27/new-papers-map-how-much-private-data-agents-leak-through-ord (subscription information: https://polylog.news/pricing).