CVE-2026-70485: Open WebUI: NAT64 IPv6 trick bypasses SSRF filter, leaks metadata

GHSA-8x5v-cpv7-8jjp HIGH PoC AVAILABLE CISA: TRACK*
Published August 4, 2026
CISO Take

Open WebUI's SSRF protection for RAG URL ingestion and web-search fetches checks whether an IPv6 address is globally routable, but never unwraps the IPv4 address hidden inside NAT64-encoded addresses — so any authenticated, non-admin user can wrap an internal IP or the cloud metadata endpoint (169.254.169.254) in the `64:ff9b::/96` prefix and read the response body back through the API. This isn't universally exploitable — it only fires on networks with a NAT64 gateway, which is common on IPv6-only or dual-stack cloud/Kubernetes deployments, hence the high attack-complexity CVSS of 7.1 and the low absolute EPSS (0.00222). But where it works, it's full-read SSRF against IAM role credentials, loopback admin panels, and internal VPC APIs, a public PoC already exists, and Open WebUI carries 150 other CVEs in the same package with 4 downstream dependents. Patch to v0.11.0 now (no config change needed), and in the meantime check whether your deployment network provides NAT64 translation — if it doesn't, this specific bug is moot for you.

Sources: NVD GitHub Advisory EPSS CISA KEV ATLAS

What is the risk?

Moderate-high risk with a narrow but common exposure window. Impact is severe if exploitable (full-read SSRF to cloud IAM credentials and internal services, CVSS 7.1, C:H), but exploitability is gated entirely on the target network providing NAT64 translation — a non-trivial precondition that many IPv4-only or classic dual-stack deployments won't meet, but one that is the default on IPv6-only cloud and Kubernetes networks (AWS EKS IPv6 clusters, GCP, many managed K8s setups). Authentication requirement is low (any verified, non-admin user), so any multi-tenant or semi-open Open WebUI instance is at risk from insiders or low-trust accounts. EPSS is low in absolute terms and the CVE is not in CISA KEV or covered by a Nuclei template, indicating no evidence of active mass exploitation yet, but a working PoC is public and the bypass technique (transition-encoding of IPv4-in-IPv6) is a known SSRF pattern that scanners will likely add.

How does the attack unfold?

Entry point
An authenticated, non-admin Open WebUI user submits a URL for RAG ingestion or web-search retrieval, targeting an internal address or cloud metadata endpoint.
AML.T0049
Filter bypass
The target IPv4 address is wrapped in the NAT64 well-known prefix (64:ff9b::/96), which passes the server's is_global() routability check because that check only inspects the literal IPv6 address.
SSRF request execution
The Open WebUI server issues the outbound GET; on a NAT64-enabled network the gateway translates the address back to the internal or metadata IPv4 target, and the request reaches it.
Impact
The internal response body — e.g., cloud IAM role credentials from the metadata service or internal API data — is returned to the requesting user through the API, enabling credential theft or internal reconnaissance.
AML.T0055

What systems are affected?

Package Ecosystem Vulnerable Range Patched
Open WebUI pip >= 0.9.0, < 0.11.0 0.11.0
152.6K 3 dependents Pushed 2d ago 83% patched ~5d to patch Full package profile →

Do you use Open WebUI? You're affected.

How severe is it?

CVSS 3.1
7.1 / 10
EPSS
0.2%
chance of exploitation in 30 days
Higher than 14% of all CVEs
Exploitation Status
Exploit Available
Exploitation: MEDIUM
Sophistication
Moderate
Exploitation Confidence
medium
CISA SSVC: Public PoC
Public PoC indexed (trickest/cve)
Composite signal derived from CISA KEV, VulnCheck KEV, CISA SSVC, EPSS, Metasploit, Exploit-DB, trickest/cve, Nuclei templates, and inthewild.io exploitation reports.

What is the attack surface?

AV AC PR UI S C I A
AV Network
AC High
PR Low
UI None
S Changed
C High
I Low
A None

What should I do?

1 step
  1. Upgrade to Open WebUI v0.11.0 or later — the fix (commit 1717b493d) unwraps IPv4 embedded in all standardized IPv6 transition encodings before the routability check, at all three checkpoints, with no configuration change required. If immediate upgrade isn't possible: disable RAG URL ingestion, URL-to-markdown, and web-search fetch features for non-trusted users; verify whether your network actually provides NAT64 translation (if not, you are not exposed by this specific vector); and block egress to 64:ff9b::/96 and other IPv6 transition prefixes at the network/firewall layer as defense-in-depth. Detection: review outbound HTTP logs from the Open WebUI server for IPv6 literals in the 64:ff9b::/96 range or other transition-encoded ranges (::ffff:0:0/96, 2002::/16), especially requests to metadata-style paths (/latest/meta-data/, /computeMetadata/).

What does CISA's SSVC say?

Decision Track*
Exploitation poc
Automatable No
Technical Impact partial

Source: CISA Vulnrichment (SSVC v2.0). Decision based on the CISA Coordinator decision tree.

How is it classified?

Which compliance frameworks are affected?

This CVE is relevant to:

EU AI Act
Article 15 - Accuracy, robustness and cybersecurity
ISO 42001
A.6.2 - AI system security controls
NIST AI RMF
MEASURE 2.7 - AI system security and resilience evaluation
OWASP LLM Top 10
LLM02:2025 - Sensitive Information Disclosure

Frequently Asked Questions

What is CVE-2026-70485?

Open WebUI's SSRF protection for RAG URL ingestion and web-search fetches checks whether an IPv6 address is globally routable, but never unwraps the IPv4 address hidden inside NAT64-encoded addresses — so any authenticated, non-admin user can wrap an internal IP or the cloud metadata endpoint (169.254.169.254) in the `64:ff9b::/96` prefix and read the response body back through the API. This isn't universally exploitable — it only fires on networks with a NAT64 gateway, which is common on IPv6-only or dual-stack cloud/Kubernetes deployments, hence the high attack-complexity CVSS of 7.1 and the low absolute EPSS (0.00222). But where it works, it's full-read SSRF against IAM role credentials, loopback admin panels, and internal VPC APIs, a public PoC already exists, and Open WebUI carries 150 other CVEs in the same package with 4 downstream dependents. Patch to v0.11.0 now (no config change needed), and in the meantime check whether your deployment network provides NAT64 translation — if it doesn't, this specific bug is moot for you.

Is CVE-2026-70485 actively exploited?

Proof-of-concept exploit code is publicly available for CVE-2026-70485, increasing the risk of exploitation.

How to fix CVE-2026-70485?

Upgrade to Open WebUI v0.11.0 or later — the fix (commit 1717b493d) unwraps IPv4 embedded in all standardized IPv6 transition encodings before the routability check, at all three checkpoints, with no configuration change required. If immediate upgrade isn't possible: disable RAG URL ingestion, URL-to-markdown, and web-search fetch features for non-trusted users; verify whether your network actually provides NAT64 translation (if not, you are not exposed by this specific vector); and block egress to `64:ff9b::/96` and other IPv6 transition prefixes at the network/firewall layer as defense-in-depth. Detection: review outbound HTTP logs from the Open WebUI server for IPv6 literals in the `64:ff9b::/96` range or other transition-encoded ranges (`::ffff:0:0/96`, `2002::/16`), especially requests to metadata-style paths (`/latest/meta-data/`, `/computeMetadata/`).

What systems are affected by CVE-2026-70485?

This vulnerability affects the following AI/ML architecture patterns: RAG pipelines, agent frameworks, model serving.

What is the CVSS score for CVE-2026-70485?

CVE-2026-70485 has a CVSS v3.1 base score of 7.1 (HIGH). The EPSS exploitation probability is 0.23%.

What is the AI security impact?

Affected AI Architectures

RAG pipelinesagent frameworksmodel serving

MITRE ATLAS Techniques

AML.T0025 Exfiltration via Cyber Means
AML.T0049 Exploit Public-Facing Application
AML.T0055 Unsecured Credentials

Compliance Controls Affected

EU AI Act: Article 15
ISO 42001: A.6.2
NIST AI RMF: MEASURE 2.7
OWASP LLM Top 10: LLM02:2025

What are the technical details?

Original Advisory

## Summary Open WebUI fetches user-supplied URLs on the server for RAG URL ingestion, URL-to-markdown conversion and web-search content retrieval, and decides whether a destination is allowed by asking whether its IP address is globally routable. That test operates on the literal IPv6 address and does not look at the IPv4 address embedded inside it. On a deployment whose network has a NAT64 gateway, any verified user can wrap an internal or cloud-metadata IPv4 address in the NAT64 well-known prefix, pass the filter, and receive the internal response body back through the API. ## Preconditions - Any verified (authenticated) user account. No admin role, no elevated permission. - Default configuration: `ENABLE_LOCAL_WEB_FETCH` off, the default `WEB_FETCH_FILTER_LIST` metadata blocklist in place. Neither prevents this, because the blocklist matches hostname strings and the NAT64 literal is not one of them. - The deployment's network must provide NAT64 translation for the well-known `64:ff9b::/96` prefix, which is the common default on IPv6-only and dual-stack cloud and Kubernetes networks. - Deployments on IPv4-only networks, or on any network without a NAT64 gateway, are not affected: the address has nowhere to route. ## Impact On an affected network a low-privilege user can read GET responses from services the server can reach but the internet cannot: cloud instance metadata including IAM role credentials, loopback-bound admin surfaces, and internal APIs in the same VPC or cluster. The response body is returned to the caller, so this is full-read, not blind. Exploitation is not universal, it depends entirely on the deployment's network providing NAT64 translation, which is why the score carries high attack complexity. Deployments without NAT64 lose nothing here. ## Fix Fixed in v0.11.0 by commit `1717b493d`. Address classification now unwraps the IPv4 embedded in IPv6 transition encodings before deciding whether a destination is global, and applies that at all three checkpoints. NAT64-wrapped public destinations continue to work. Upgrading to v0.11.0 fully resolves the issue with no configuration change. ## Root cause - `backend/open_webui/retrieval/web/utils.py` — `validate_url()`, the pre-fetch check on the submitted URL. - `backend/open_webui/retrieval/web/utils.py` — `_ssrf_safe_new_conn()` and `_SSRFSafeResolver`, the connect-time re-checks that defeat DNS rebinding. All three decided reachability from `ipaddress.ip_address(ip).is_global` applied to the literal address. That predicate answers whether an IPv6 address sits in globally-routable space, which is a different question from where the packet actually ends up once a transition gateway translates it. The NAT64 well-known prefix is by design a global prefix carrying an arbitrary IPv4 destination, so an internal target wrapped in it satisfies the check while reaching exactly what the check exists to prevent. Because the same predicate backed the connect-time re-checks, no later layer caught it either. The fix inspects every standardized transition encoding rather than only the NAT64 prefix, since the same reasoning error applies to each of them. ## Proof of concept Against the real `POST /api/v1/retrieval/process/web` flow on v0.10.2 as an authenticated user, with internal HTTP services returning a marker string. The plain forms are rejected with HTTP 400: ``` http://169.254.169.254/latest/meta-data/ -> 400 http://127.0.0.1/ -> 400 http://[::ffff:169.254.169.254]/ -> 400 http://metadata.google.internal/ -> 400 ``` The NAT64 encodings of the same targets are accepted, and the response body is returned in the `content` field: ``` http://[64:ff9b::a9fe:a9fe]/latest/meta-data/iam/security-credentials/ -> 200, marker returned http://[64:ff9b::7f00:1]/admin/internal-status -> 200, marker returned ``` NAT64 translation was modelled by binding the translated addresses locally rather than by routing through a real NAT64 gateway; everything else, including the request flow and the validation code, is the unmodified v0.10.2 path. After the fix both URLs return 400 while `http://[64:ff9b::808:808]/` (8.8.8.8, public) still returns 200, confirming no over-blocking. ## Credits - tonghuaroot — reported the transition-form gap in the address classification and supplied the fix approach.

Exploitation Scenario

A contractor or low-trust employee with a standard (non-admin) Open WebUI account submits a document or chat message referencing `http://[64:ff9b::a9fe:a9fe]/latest/meta-data/iam/security-credentials/` for RAG ingestion. The server-side pre-fetch validator sees a globally-routable IPv6 literal and allows it; on the organization's NAT64-enabled cloud network, the gateway translates the address back to 169.254.169.254 and the request reaches the real cloud metadata service. The IAM role credentials returned in the response body are rendered back to the user inside the chat/RAG content field, handing them temporary cloud credentials scoped to whatever role the Open WebUI instance runs under — which the attacker can then use to enumerate or access other cloud resources, well outside anything the chat application was meant to expose.

Weaknesses (CWE)

CWE-918 — Server-Side Request Forgery (SSRF): The web server receives a URL or similar request from an upstream component and retrieves the contents of this URL, but it does not sufficiently ensure that the request is being sent to the expected destination.

Source: MITRE CWE corpus.

CVSS Vector

CVSS:3.1/AV:N/AC:H/PR:L/UI:N/S:C/C:H/I:L/A:N

Timeline

Published
August 4, 2026
Last Modified
August 5, 2026
First Seen
August 4, 2026

Related Vulnerabilities