CVE-2026-61426 describes a vulnerability in MervinPraison's PraisonAI prior to version 1.7.3 where an insecure default configuration allows unauthenticated remote attackers to access sensitive information and invoke agent functionality. The application binds to all interfaces, lacks API key enforcement, and permits wildcard CORS, enabling calls to GET /api/agents to read agent instructions and system prompts, and POST /api/chat to invoke agents without authentication. No official patch or remediation guidance is currently documented.

Unauthenticated attackers can retrieve sensitive agent instructions and system prompts, potentially exposing confidential operational details. They can also invoke agents remotely without authorization, which may lead to unauthorized actions or data manipulation within the affected system. The vulnerability does not require user interaction and can be exploited remotely over the network.

Patch status is not yet confirmed — check the vendor advisory for current remediation guidance. Until an official fix is available, users should avoid exposing PraisonAI instances to untrusted networks or restrict network access to trusted hosts only. Additionally, manually configuring API key requirements and restricting CORS policies may reduce exposure.