CVE-2026-61539
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.
Does this matter?
High impact if exploited, but EPSS currently rates exploitation as unlikely (1.23%). Schedule it in the normal patch cycle and watch for a rise in EPSS or a public exploit.
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.
- CVSS 3.1
- 10.0 CRITICALCVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:H/A:H
- EPSS
- 1.23% probability · 69th percentile
- Public exploits
- None in Exploit-DB
- CISA KEV
- Not listed
- Weakness
- CWE-95
- Source
- security-advisories@github.com
EPSS 1.2% since 27 October 2026; no change recorded yet. Points are recorded when the score first appears and whenever it moves by a percentage point or more.
References
Source: NVD record, EPSS from FIRST.org, KEV from CISA, exploits from Exploit-DB. Refreshed daily. Download this record as JSON.