If you do one security check this quarter, make it agent memory
Vectorize CEO Chris Latimer says coding-agent memory keeps API keys in plaintext and can be poisoned via plugins and MCP.
Chris Latimer, CEO of Vectorize, told Help Net Security that coding-agent memory often keeps API keys, credentials, and sensitive documents in plaintext on developer machines, cloud memory services, and markdown files. He said attackers can plant poisoned memories through plugins, skills, and MCP integrations, including lures aimed at inexperienced users of tools such as Claude Code. Access control for agent memory generally lags RBAC and ABAC. He urged CISOs to audit which memory products are in use and cited the OWASP Memory Guard project for filtering content before it is persisted.
- Coding agents persist API keys, credentials, and documents in plaintext memory.
- Poisoned memories can be planted through plugins, skills, and MCP integrations.
- Inexperienced users are lured by too-good-to-be-true coding-agent plugins.
- Agent-memory access control lags RBAC and ABAC used on other data.
- CISOs should audit memory tools for ungoverned use and leaked secrets.
- Products
- Claude CodeMemory Guard
From the original article
4 min read at helpnetsecurity.com
In this interview with Help Net Security, Chris Latimer, CEO of Vectorize, talks about the security risks hiding in AI agent memory. He found coding agents storing API keys, credentials, and sensitive documents in plain text on developer machines and in cloud services.
By Mirko Zorz · helpnetsecurity.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.