Executive brief
9Router, an AI model routing dashboard, contains multiple security flaws that allow unauthorized individuals to access sensitive data. Attackers can view private AI conversation histories, including user messages and system prompts, and harvest plaintext API keys for connected AI services. This could lead to the theft of proprietary AI instructions, exposure of private user communications, and unauthorized use of paid AI accounts.
Technical details
Multiple Next.js API routes in 9Router (specifically under /api/providers and /api/usage) lack authentication middleware. A remote, unauthenticated attacker can perform CRUD operations on provider connections, retrieve full plaintext API keys via /api/usage/stats, and enumerate request logs. Furthermore, the /api/usage/request-details endpoint allows for the retrieval of complete AI conversation histories, including system prompts, tool calls, and user email addresses. The vulnerability stems from a failure to implement authorization checks and sensitive data masking in the dashboard's backend routes.
Affected products
- decolua 9Router <= 0.4.41
Timeline
- 2026-06-13: advisory: Initial GitHub Security Advisory published by vendor
- 2026-07-13: disclosed: CVE published and NVD record created