CVE-2026-61539

Name of the Vulnerable Software and Affected Versions Xinference versions prior to 2.7.0

Description Xinference passes attacker-influenced Llama3 tool-call output to the eval() function within xinference/model/llm/tool parsers/llama3 tool parser.py and xinference/model/llm/utils.py. Requests sent to the '/v1/chat/completions' endpoint containing a tools field are processed through xinference/api/restful api.py, xinference/model/llm/transformers/core.py, handle chat result non streaming(), and post process completion() before the extract tool calls() or eval llama3 chat arguments() functions evaluate the model-generated Python expression. An unauthenticated remote attacker can use a crafted prompt to influence this output and execute arbitrary commands within the context of the Xinference server process.

Recommendations Update to version 2.7.0.