CVE-2026-61539: CWE-95: Improper Neutralization of Directives in Dynamically Evaluated Code ('Eval Injection') in xorbitsai inference
CVE-2026-61539 is a critical remote code execution vulnerability in xorbitsai's Xinference inference API. Versions 2.5.0 and earlier improperly pass attacker-controlled output from Llama3 tool calls to Python's eval() function, allowing unauthenticated attackers to execute arbitrary code on the server. The flaw is fixed in version 2.7.0.
AI Analysis
Technical Summary
Xinference versions 2.5.0 and earlier contain an eval injection vulnerability (CWE-95) where attacker-influenced output from Llama3 tool calls is passed directly to Python's eval() function in specific modules (llama3_tool_parser.py and utils.py). Requests to the /v1/chat/completions endpoint with a tools field flow through several internal functions before the model-generated Python expression is evaluated. This allows unauthenticated remote attackers to execute arbitrary commands in the server process context. The issue is resolved in version 2.7.0.
Potential Impact
An unauthenticated remote attacker can execute arbitrary code on the Xinference server, leading to full compromise of confidentiality, integrity, and availability of the affected system. The CVSS score is 10.0 (critical), indicating maximum impact with network attack vector, no privileges required, and no user interaction needed.
Mitigation Recommendations
Upgrade Xinference to version 2.7.0 or later, where this vulnerability is fixed. No other official remediation or temporary mitigations are provided. Patch status is not explicitly confirmed in vendor advisory content, but the fix version is stated.
CVE-2026-61539: CWE-95: Improper Neutralization of Directives in Dynamically Evaluated Code ('Eval Injection') in xorbitsai inference
Description
CVE-2026-61539 is a critical remote code execution vulnerability in xorbitsai's Xinference inference API. Versions 2.5.0 and earlier improperly pass attacker-controlled output from Llama3 tool calls to Python's eval() function, allowing unauthenticated attackers to execute arbitrary code on the server. The flaw is fixed in version 2.7.0.
CVSS v3.1
Score 10.0critical
Affected software
pkg:github/xorbitsai/inferenceRun on your own infrastructure? Check whether these packages are installed with threat-finder — our free open-source scanner.
Weaknesses
AI-Powered Analysis
Machine-generated threat intelligence
Technical Analysis
Xinference versions 2.5.0 and earlier contain an eval injection vulnerability (CWE-95) where attacker-influenced output from Llama3 tool calls is passed directly to Python's eval() function in specific modules (llama3_tool_parser.py and utils.py). Requests to the /v1/chat/completions endpoint with a tools field flow through several internal functions before the model-generated Python expression is evaluated. This allows unauthenticated remote attackers to execute arbitrary commands in the server process context. The issue is resolved in version 2.7.0.
Potential Impact
An unauthenticated remote attacker can execute arbitrary code on the Xinference server, leading to full compromise of confidentiality, integrity, and availability of the affected system. The CVSS score is 10.0 (critical), indicating maximum impact with network attack vector, no privileges required, and no user interaction needed.
Mitigation Recommendations
Upgrade Xinference to version 2.7.0 or later, where this vulnerability is fixed. No other official remediation or temporary mitigations are provided. Patch status is not explicitly confirmed in vendor advisory content, but the fix version is stated.
Technical Details
- Data Version
- 5.2
- Assigner Short Name
- GitHub_M
- Date Reserved
- 2026-07-10T16:27:03.093Z
- Cvss Version
- 3.1
- State
- PUBLISHED
- Remediation Level
- null
Threat ID: 6a88baa8acd9273b49b84378
Added to database: 08/21/2026, 20:52:56 UTC
Last enriched: 08/21/2026, 21:07:13 UTC
Last updated: 08/21/2026, 21:32:10 UTC
Views: 4
Community Reviews
0 reviewsCrowdsource mitigation strategies, share intel context, and vote on the most helpful responses. Sign in to add your voice and help keep defenders ahead.
Want to contribute mitigation steps or threat intel context? Sign in or create an account to join the community discussion.
Actions
Updates to AI analysis require Pro Console access. Upgrade inside Console → Billing.
Need more coverage?
Upgrade to Pro Console for AI refresh and higher limits.
For incident response and remediation, OffSeq services can help resolve threats faster.
Latest Threats
Check if your credentials are on the dark web
Instant breach scanning across billions of leaked records. Free tier available.