Data poisoning and RAG manipulation
Summary
A discussion of how data poisoning and RAG manipulation pose a silent, dangerous threat to AI systems, arguing that security must extend beyond input filtering to memory, data pipelines, and multi-agent logic.
Similar Articles
How Advanced Prompt Injections Hijack AI Agents (And Why Basic Filters Aren’t Enough)
Article warns that basic content filters are insufficient against advanced prompt injection attacks on AI agents, especially in RAG pipelines, and calls for robust input sanitization and architectural defenses.
Towards Safer RAG: Only Agents Capable of System 2 Thinking may Access Untrusted Documents
This paper proposes that only AI agents capable of System 2 thinking should access untrusted documents in RAG systems to enhance security, introducing new metrics to evaluate robustness and showing reasoning models are more resistant to knowledge poisoning.
we keep talking about making agents smarter but not about making them safe around data
The article argues that AI agent safety focuses too much on instruction-following and not enough on data access governance, highlighting the Agentic Data Protocol as an early effort to put policy in infrastructure.
The Containment Gap: How Deployed Agentic AI Frameworks Fail Public-Facing Safety Requirements
This paper audits LangChain, AutoGPT, and OpenAI Agents SDK for architectural safety guarantees and finds no native compliance with containment principles, demonstrating that memory poisoning can cause persistent failures; it introduces lightweight mechanisms to eliminate such attacks.
Are we underestimating how dangerous agent memory can become?
Discusses the risks of giving AI agents memory, including trust issues, data poisoning, and operational risks, and poses key questions for builders.