CVE-2026-57127

PraisonAI, an open-source multi-agent framework, fails to enforce authentication in its 'recipe serve' component before version 4.6.58. When an operator selects api-key or JWT authentication, the APIKeyAuthMiddleware or JWTAuthMiddleware incorrectly forwards requests when PRAISONAI_API_KEY or PRAISONAI_JWT_SECRET (and the corresponding recipe value) are absent, so the server silently runs without the intended protection. An unauthenticated network attacker who can reach the service can invoke recipe execution, input, and output surfaces and may trigger connected tools, potentially reading sensitive agent data and driving any integrated actions the agents can perform. Operators running 'recipe serve' on any version before 4.6.58 with authentication enabled are affected, with the highest risk where the service is exposed to untrusted networks. The issue is fixed in 4.6.58; no public proof of concept or observed exploitation is known at this time.

What to do: Upgrade PraisonAI to 4.6.58 or later, which makes the middleware fail closed when authentication is enabled. As defense in depth, explicitly set PRAISONAI_API_KEY or PRAISONAI_JWT_SECRET and the recipe-level auth values, and restrict network access to recipe serve endpoints. Review logs on any pre-4.6.58 deployment for unauthenticated requests to recipe execution, input, or output endpoints, and audit what connected tools those recipes could have triggered.