Microsoft Warns Poisoned MCP Tool Descriptions Can Make AI Agents Leak Data

A single invisible comment in an Azure DevOps pull request can turn a reviewer's own AI coding agent against them, driving it into projects the attacker has no rights to reach and

Ask an AI coding agent to scan open-source code for security holes, and it might run the attacker's code on your own machine instead. That is the finding in

A malicious tool server connected to an AI coding assistant can quietly walk off with SSH keys, environment secrets, source code, and customer data without ever sending one

<p><a class="article-body-image-wrapper"

Ask an AI agent to summarize the reviews on a product page, and a single planted review can make it click "Buy Now" instead. Ask a coding assistant to apply a maintainer's fix

Microsoft wants enterprises to closely watch the agents employees use with systems not controlled by AI labs like OpenAI and Anthropic.

Cybersecurity researchers have described what they say is a new class of attack that can trick artificial intelligence (AI) coding agents into running arbitrary code on developer

<p>WebMCP gives AI agents named tools to call, and a clean route for prompt injection. Here's what Chrome says to lock down first.</p> <p>The post <a