{"apiVersion":"1.0","identifier":"CVE-2026-61539","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.","publishedAt":"2026-08-21T21:17:00","lastModifiedAt":"2026-08-24T20:16:52","sourceUrl":"https://nvd.nist.gov/vuln/detail/CVE-2026-61539","cvssScore":10,"cvssVector":"CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:H/A:H","epssProbability":0.00657,"riskScore":1.06,"affectedProduct":"Xinference","affectedVersions":"<=2.5.0","vulnerabilityType":"Web app","operatingSystems":[],"links":{"self":"https://www.redsauce.net/api/cves/CVE-2026-61539","webPages":{"es":"https://www.redsauce.net/es/cves/CVE-2026-61539","en":"https://www.redsauce.net/en/cves/CVE-2026-61539","fr":"https://www.redsauce.net/fr/cves/CVE-2026-61539","pt":"https://www.redsauce.net/pt/cves/CVE-2026-61539","de":"https://www.redsauce.net/de/cves/CVE-2026-61539","sk":"https://www.redsauce.net/sk/cves/CVE-2026-61539","el":"https://www.redsauce.net/el/cves/CVE-2026-61539"},"source":"https://nvd.nist.gov/vuln/detail/CVE-2026-61539"}}