CVE-2026-56663: Significant-Gravitas AutoGPT SSRF bypass in SendWebRequestBlock

CVE-2026-56663 · Severity: high · CVSS 8.5 · Published 2026-06-26

AutoGPT is a platform used to create and manage autonomous AI agents. A security flaw allows authenticated users to bypass network restrictions and force the platform to connect to internal services that should be private, such as cloud metadata endpoints or internal databases. This could lead to the theft of sensitive internal data or unauthorized access to other systems within the organization's private network.

Technologies: Significant-Gravitas AutoGPT. Vendors: Significant-Gravitas.

Quick answers

Is CVE-2026-56663 exploited in the wild?
No. Junglewise Threat Intelligence has no report of CVE-2026-56663 being exploited in the wild.
How severe is CVE-2026-56663?
High severity, with a CVSS score of 8.5, published 26 June 2026.
Which products and versions are affected?
Significant-Gravitas AutoGPT < 0.6.52.
Is there a fix for CVE-2026-56663?
A patch was published on 2026-06-26: Fix released in version 0.6.52 Junglewise finds every place you run it, drafts the upgrade and applies it once you approve.
How is CVE-2026-56663 attacked?
It is attacked over the network, in a software library.
Am I affected by CVE-2026-56663?
Check whether you run Significant-Gravitas AutoGPT at an affected version. Junglewise does that check for you around the clock: connect your code, cloud and domains, and it matches new vulnerabilities like this one against what you actually run and tells you when one reaches you.

Technical details

A Server-Side Request Forgery (SSRF) vulnerability exists in the SendWebRequestBlock component of AutoGPT. The validation function _is_ip_blocked() in backend/backend/util/request.py fails to normalize IPv4-mapped IPv6 addresses (e.g., ::ffff:a.b.c.d) before comparing them against IPv4 blocklists. Because Python's ipaddress module treats these as distinct types, the internal IPv4 address bypasses the filter. Additionally, the blocklist lacks special-use ranges like 100.64.0.0/10 (CGNAT). An attacker can exploit this by providing a hostname that resolves to an IPv4-mapped IPv6 address, allowing them to probe or interact with internal-only network services. The issue is fixed in version 0.6.52 by implementing proper IP normalization and expanding the blocklist.

Affected products

  • Significant-Gravitas AutoGPT < 0.6.52

Timeline

  • 2026-06-24: advisory: GitHub Security Advisory published
  • 2026-06-26: disclosed: CVE published to NVD
  • 2026-06-26: patched: Fix released in version 0.6.52

References

This page as Markdown. Free to reuse under CC BY 4.0 with a link back.