Safeguard
Vulnerability Analysis•

CVE-2026-40114: SSRF via webhook_url in PraisonAI

CVE-2026-40114 is a server side request forgery in the PraisonAI jobs API before 4.5.128. Upgrade to 4.5.128 and keep the jobs API off untrusted networks.

Safeguard Security Team
2 min read

CVE-2026-40114 is a server side request forgery flaw in PraisonAI, scored CVSS 7.2. Before version 4.5.128, the /api/v1/runs endpoint accepts any webhook_url and posts job results to it, with no authentication. Upgrade to 4.5.128 or later and keep the jobs API off untrusted networks.

What's affected

  • Affected: PraisonAI before 4.5.128
  • Fixed: 4.5.128

The advisory says the Jobs API server has no authentication middleware and allows requests from any origin. The webhook_url field is a plain string with no scheme or host checks, and the server sends an HTTP POST to it when a job finishes or fails.

How to check if you're exposed

Check the installed version:

pip show praisonai

Then find out whether the jobs API server is running anywhere and who can reach it. If port access is open beyond localhost, treat it as exposed until you upgrade. Check outbound firewall logs from the host for POST requests to internal addresses or to cloud metadata addresses you did not expect.

What to do now

  1. Upgrade PraisonAI to 4.5.128 or later. Moving to 4.5.133 also covers the related fixes in this series.
  2. Put the jobs API behind authentication or a reverse proxy that requires it.
  3. Block outbound traffic from the host to internal ranges and metadata services unless the job runner needs it.
  4. Review logs for webhook calls to unusual destinations.

Why it matters

The weakness is CWE-918, server side request forgery. The server makes a request on behalf of whoever submits a job, and it trusts the destination. On a cloud host that can mean reaching internal services or metadata endpoints that the caller could not reach directly.

The advisory notes the check is missing at every step: no scheme restriction, no private address filtering, no authentication in front. Any one of these would have limited the impact. The record does not flag exploitation in the wild.

Quick check

Is the payload sensitive?

The advisory shows the POST body carries the job id, status, and the job result or error, so results could go to the destination a caller chooses.

Sources

Last reviewed: 19 October 2026.

Never miss an update

Weekly insights on software supply chain security, delivered to your inbox.

Self-healing security runs on Safeguard.

Your first fix PR is minutes away.

No sales call required, even your agent can complete the purchase over MCP.