Vulnerability · Published Aug 21, 2026 · Assigned by GitHub_M
CVE-2026-61539Xinference: Remote code execution via unsafe `eval()` in Llama3 tool-call parsing
Critical 10.0
In plain English
A critical-severity security flaw in inference from xorbitsai. There are no confirmed reports of exploitation so far. The chance it is exploited in the next 30 days is estimated at 1.2% (higher than 69% of all known vulnerabilities). Fixed versions are available - upgrade the affected package.
Official description
Xinference is an inference API for running open-source, speech, and multimodal models. In 2.5.0 and earlier, Xinference passes attacker-influenced Llama3 tool-call output to eval() in xinference/model/llm/tool_parsers/llama3_tool_parser.py and xinference/model/llm/utils.py. Requests to /v1/chat/completions with a tools field flow through xinference/api/restful_api.py, xinference/model/llm/transformers/core.py, handle_chat_result_non_streaming(), and _post_process_completion() before extract_tool_calls() or _eval_llama3_chat_arguments() evaluates the model-generated Python expression. An unauthenticated remote attacker can influence that output through a crafted prompt and execute commands in the Xinference server process context. This issue is fixed in version 2.7.0.
Affected packages & fixed versions
| Ecosystem | Package | Fixed in | Advisory |
|---|---|---|---|
| PyPI | xinference | 2.7.0 | GHSA-x2rj-828p-hx9m |
| PyPI | xinference | 2.7.0 | PYSEC-2026-3946 |
Affected products (as reported)
Weakness: CWE-95 · Vector: CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:H/A:H
References
- https://github.com/xorbitsai/inference/security/advisories/GHSA-x2rj-828p-hx9m x_refsource_CONFIRM
- https://github.com/xorbitsai/inference/pull/4786 x_refsource_MISC
- https://github.com/xorbitsai/inference/commit/1b3d220f342ce68d34cec4586d9409d457dadc42 x_refsource_MISC
- https://github.com/xorbitsai/inference/releases/tag/v2.7.0 x_refsource_MISC