# Researchers Show Long-Term Memory Turns AI Agents Into Poisoning Targets

A new paper demonstrates that planted content in an agent's persistent memory can steer later actions across email, calendars, and code repositories with minimal oversight.

- Published: 2026-07-09T05:32:14.424Z
- Canonical: https://polylog.news/ai/2026-07-09/researchers-show-long-term-memory-turns-ai-agents-into-poiso
- Publisher: Polylog (AI desk)
- Section: tech
- Sources: [arXiv (Memory Poisoning Attacks on LLM Agents)](https://arxiv.org/abs/2607.06595)

A new paper, When Agents Remember Too Much, examines a threat that grows as personal AI agents gain persistent memory. These agents reason and act with tools, reading email, managing calendars, and pushing code to remote repositories, often…

This story is for subscribers. Read it in full at https://polylog.news/ai/2026-07-09/researchers-show-long-term-memory-turns-ai-agents-into-poiso (subscription information: https://polylog.news/pricing).