Glow Labs reveals AI coding agents installing deprecated MCP servers on their own
Understand this story
The Glow Labs research team found that AI coding agents, when users request a certain capability, install deprecated MCP servers on their own, a phenomenon they call MCPhantom. These servers no longer receive security updates and may expose organizations to high-severity vulnerabilities. Glow Labs says that in 98% of sessions where agents installed deprecated MCP servers, it was because the package had appeared in model training data—for example, the agent chose Anthropic's archived GitHub server rather than the officially maintained GitHub version; in the remaining 2%, the user explicitly asked for the deprecated package and the agent complied. Glow Labs says no one has yet been seen exploiting these vulnerabilities. Glow Labs says its customers can find all deprecated MCP servers on a developer's machine with a single request.
AI overview based on source reports · Updated 2 days ago
Coverage timeline
Follow the coverage to explore different perspectives.
- Hacker News AIAI coding agents are installing abandoned MCP servers with known security issues
Glow Labs found that AI coding agents often install deprecated MCP servers because those package names appeared in their training data, a pattern it calls MCPhantom; in 98% of such sessions the agent chose the retired server on its own rather than because the user named it. The team identified four high-severity vulnerabilities, including a remote code execution flaw in the Puppeteer server, across Anthropic's archived reference servers, which have 17M npm downloads since July 2025 and will receive no fixes. Because these npx-launched servers appear in no dependency manifest, EDR, SCA, harness guardrails and software inventory tools miss them, leaving the endpoint as the place to detect and remove them.
Activity for this story
- Current comparable activity
- 3
- Peak comparable activity
- 10Nov 4 20:00
- Change in the past 24 hours
- -49%
Trends compare the same fully observed participants, which may cover fewer participants than current activity. Move the pointer or tap the chart to inspect hourly activity; use the left and right arrow keys with a keyboard.