# CVE-2026-85668: Xinference arbitrary file read in LLM auto-register endpoint

Source: https://junglewise.ai/threats/cve-2026-85668-xinference-arbitrary-file-read-in-llm-auto-register-endpoint  
Publisher: Junglewise Threat Intelligence, https://junglewise.ai/threats  
Published: 2026-09-04. Updated: 2026-09-26  
License: CC BY 4.0 with a link back

CVE-2026-85668 · Severity: high · CVSS 7.5 · Published 2026-09-04

Xinference is an inference server that allows running open-source language models through a unified API. An unauthenticated attacker can exploit a file read vulnerability in the model auto-register endpoint to read arbitrary files from the server, potentially exposing sensitive configuration data and other files stored on the same filesystem.

## Quick answers

**Is CVE-2026-85668 exploited in the wild?** No. Junglewise Threat Intelligence has no report of CVE-2026-85668 being exploited in the wild.

**How severe is CVE-2026-85668?** High severity, with a CVSS score of 7.5, published 4 September 2026.

**Which products and versions are affected?** Xorbits Xinference v3.x (affected commit 4a94832).

**Is there a fix for CVE-2026-85668?** No fixed version is recorded yet. Check the vendor's advisory in the references below. If you run it, Junglewise drafts a mitigation that fits your setup and applies it once you approve, with senior security engineers behind it.

**How is CVE-2026-85668 attacked?** It is attacked over the network, in a hosted service.

**Am I affected by CVE-2026-85668?** Check whether anything you run is named in the advisory. 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

The vulnerability is an unauthenticated arbitrary-path file read flaw in the POST /v1/models/llm/auto-register endpoint. The endpoint accepts a caller-supplied model_path parameter without authentication or path validation. An attacker can supply a path to any directory, and the endpoint will read and parse config.json, tokenizer_config.json, and chat_template.jinja files at that location, reflecting the parsed content back to the caller. This allows an unauthenticated attacker to probe the server filesystem and extract file contents from any directory accessible by the Xinference process.

## Affected products

- Xorbits Xinference v3.x (affected commit 4a94832)

## Timeline

- 2026-09-04: disclosed

## References

- [https://github.com/xorbitsai/inference](https://github.com/xorbitsai/inference)
- [https://github.com/xorbitsai/inference/blob/v3.3.0/xinference/model/llm/config_parser.py](https://github.com/xorbitsai/inference/blob/v3.3.0/xinference/model/llm/config_parser.py)