CVE-2026-63086

text-generation-inference 3.3.7 SSRF via fetch_image in multimodal chat completions

Published Jul 16, 2026 CVSS 4.0 · 6.9 Medium EPSS < 1% KEV · Not listed

Description

text-generation-inference through 3.3.7 contains a server-side request forgery (SSRF) vulnerability in the OpenAI-compatible multimodal chat completions endpoint that allows unauthenticated network attackers to coerce the server into issuing arbitrary HTTP GET requests by supplying a crafted image_url value in chat message content. The fetch_image function in router/src/validation.rs performs no validation of private, loopback, link-local, or cloud metadata target addresses, and the reqwest HTTP client follows redirects by default, enabling attackers to bypass scheme checks via redirect chains to reach internal services and cloud instance-metadata endpoints for internal port scanning and credential theft.

Analysis

Technical impact

The vulnerability is a server‑side request forgery (SSRF) in the multimodal chat completions endpoint of text‑generation‑inference 3.3.7. The fetch_image function accepts an arbitrary image_url without validating that the URL targets an external, non‑internal resource, and the HTTP client follows redirects by default, allowing attackers to bypass scheme checks through redirect chains. An unauthenticated attacker can coerce the server into issuing arbitrary HTTP GET requests, potentially accessing internal services or cloud metadata endpoints for credential harvesting.

Affected Systems

The affected product is Huggingface's text‑generation‑inference server. Versions up to and including 3.3.7 that expose the multimodal chat completions endpoint are vulnerable. No other vendors or product variants are indicated.

Risk and Exploitability

The flaw carries a CVSS score of 6.9. An attacker with network reachability can target the vulnerable server to perform internal network discovery or credential theft. The EPSS score is <1%, indicating a very low exploitation probability. The vulnerability is not listed in the CISA KEV catalog, so it is not currently a known exploited vulnerability.

Affected Products

Default status is the baseline for the product, each version can override it (e.g. patched versions marked unaffected).

Vendor Product Default status Versions

huggingface

text-generation-inference

affected

Version	Status	Constraints
`0`	affected	≤ 3.3.7

No data.

Vendor Product Versions

Huggingface

Text-generation-inference

Version	Status	Scheme	Platform
`[0,3.3.7]`	affected	generic	—

Remediation

No solution or workaround provided in the CVE record.

OpenCVE Recommended Actions

Upgrade to the latest release of text‑generation‑inference that contains the SSRF fix
Block outbound HTTP/HTTPS connections from the server to internal IP ranges, loopback, link‑local, and cloud metadata addresses by configuring firewalls or network segmentation
Implement validation to reject image_url values that resolve to private, loopback, or link‑local addresses, or use a proxy that sanitizes URLs before forwarding requests

Generated by OpenCVE AI on July 31, 2026 at 01:47 UTC.

Tracking

Advisories

No advisories available

Metrics

Attack Vector Network

Attack Complexity Low

Privileges Required None

Attack Requirements Present

User Interaction None

Vulnerable System Confidentiality Impact None

Vulnerable System Integrity Impact None

Vulnerable System Availability Impact None

Subsequent System Confidentiality Impact High

Subsequent System Integrity Impact None

Subsequent System Availability Impact None

Attack Vector Network

Attack Complexity Low

Privileges Required None

Scope Changed

Confidentiality Impact High

Integrity Impact None

Availability Impact None

User Interaction None

No CVSS v3.0

No CVSS v2

This CVE is not in the KEV list.

Exploit prediction score: < 1% (0.434% probability within 30 days)

Exploitation poc

Automatable yes

Technical Impact partial

References

Link	Providers
https://github.com/geo-chen/oss/blob/main/text-generation-inference.md
https://www.vulncheck.com/advisories/text-generation-inference-ssrf-via-fetch-image-in-multimodal-chat-completions

History

Jul 16, 2026 6:30 PM Scoring metrics updated View changes

Type	Values Removed	Values Added
Metrics		ssvc `{'options': {'Automatable': 'yes', 'Exploitation': 'poc', 'Technical Impact': 'partial'}, 'version': '2.0.3'}`

Jul 16, 2026 5:45 PM First appearance in the feed. Affected vendors updated View changes

Type	Values Removed	Values Added
First Time appeared		Huggingface Huggingface text-generation-inference
Vendors & Products		Huggingface Huggingface text-generation-inference

Jul 16, 2026 4:30 PM Description updated. Title updated. Weaknesses updated. Added references. Scoring metrics updated View changes

Type	Values Removed	Values Added
Description		text-generation-inference through 3.3.7 contains a server-side request forgery (SSRF) vulnerability in the OpenAI-compatible multimodal chat completions endpoint that allows unauthenticated network attackers to coerce the server into issuing arbitrary HTTP GET requests by supplying a crafted image_url value in chat message content. The fetch_image function in router/src/validation.rs performs no validation of private, loopback, link-local, or cloud metadata target addresses, and the reqwest HTTP client follows redirects by default, enabling attackers to bypass scheme checks via redirect chains to reach internal services and cloud instance-metadata endpoints for internal port scanning and credential theft.
Title		text-generation-inference 3.3.7 SSRF via fetch_image in multimodal chat completions
Weaknesses		CWE-918
References		https://github.com/geo-chen/oss/blob/main/text-generation-inference.md https://www.vulncheck.com/advisories/text-generation-inference-ssrf-via-fetch-image-in-multimodal-chat-completions
Metrics		cvssV3_1 `{'score': 8.6, 'vector': 'CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:N/A:N'}` cvssV4_0 `{'score': 6.9, 'vector': 'CVSS:4.0/AV:N/AC:L/AT:P/PR:N/UI:N/VC:N/VI:N/VA:N/SC:H/SI:N/SA:N'}`