Raw vector
CVSS:3.1/AV:N/AC:H/PR:L/UI:R/S:U/C:H/I:H/A:HSummary
CVE-2025-66448 is a high-severity Code Injection (CWE-94) vulnerability in Vllm Vllm. Its CVSS base score is 7.1 (High).
Operationally, exploitation aligns with the MITRE ATT&CK technique Command and Scripting Interpreter (T1059); ranked at the 49th percentile by exploit likelihood (below the median); it is not currently listed in the CISA KEV catalog.
This vulnerability is AI-related — categorised as NLP and Transformers; in the Supply Chain and Deployment risk domain.
The strongest mitigations our analysis identified map to SA-11 (Developer Testing and Evaluation) and SI-10 (Information Input Validation) — see the control section below for these in your framework.
Deeper analysis AI-assisted summary
Synthesised by an AI model from the NVD description and linked references — a reading aid, not an authoritative source.
CVE-2025-66448 is a remote code execution vulnerability in vLLM, an inference and serving engine for large language models, affecting versions prior to 0.11.1. The issue resides in the Nemotron_Nano_VL_Config class, where loading a model configuration containing an auto_map entry triggers resolution via get_class_from_dynamic_module, which fetches and instantiates Python code from a remote repository specified in the auto_map string. This execution occurs even when the caller explicitly sets trust_remote_code=False in vllm.transformers_utils.config.get_config, bypassing intended security controls.
An attacker can exploit this by publishing a seemingly benign frontend repository with a config.json file that includes an auto_map pointing to a separate malicious backend repository. A victim loading the frontend model config will silently fetch and execute the backend's arbitrary Python code on their host. Per the CVSS v3.1 score of 7.1 (AV:N/AC:H/PR:L/UI:R/S:U/C:H/I:H/A:H), exploitation requires network access, high attack complexity, low privileges, and user interaction, but yields high impacts on confidentiality, integrity, and availability, classified under CWE-94 (code injection).
The vulnerability is addressed in vLLM version 0.11.1. Official mitigation details are available in the project's security advisory at GHSA-8fr4-5q9j-m8gm, the fixing pull request at github.com/vllm-project/vllm/pull/28126, and the commit ffb08379d8870a1a81ba82b72797f196838d0c86, which practitioners should review for patch implementation guidance.
This flaw highlights risks in AI/ML inference engines handling untrusted model configurations from remote sources, with no reported real-world exploitation as of the CVE publication on 2025-12-01.
OWASP Top 10 for Web (2025)
EU & UK References
- 🇪🇺 ENISA EUVD: EUVD-2025-200115
Vulnerability Data
vLLM is an inference and serving engine for large language models (LLMs). Prior to 0.11.1, vllm has a critical remote code execution vector in a config class named Nemotron_Nano_VL_Config. When vllm loads a model config that contains an auto_map entry,…
more
the config class resolves that mapping with get_class_from_dynamic_module(...) and immediately instantiates the returned class. This fetches and executes Python from the remote repository referenced in the auto_map string. Crucially, this happens even when the caller explicitly sets trust_remote_code=False in vllm.transformers_utils.config.get_config. In practice, an attacker can publish a benign-looking frontend repo whose config.json points via auto_map to a separate malicious backend repo; loading the frontend will silently run the backend’s code on the victim host. This vulnerability is fixed in 0.11.1.
- CWE(s)
AI Security AnalysisAI
- AI Category
- NLP and Transformers
- Risk Domain
- Supply Chain and Deployment
- OWASP Top 10 for LLMs 2025
- None mapped
- Classification Reason
- Matched keywords: llms, vllm
Related Threats
MITRE ATT&CK Enterprise Techniques
CVEs Like This One
Affected Assets
Mitigating Controls
Control response
—
—
—
V1.3.1
Mitigating Controls (NIST 800-53 r5) AI
Developer testing and evaluation finds code paths that accept and execute externally influenced strings.
Input validation directly stops untrusted data from being used to construct executable code without neutralization.
Least privilege limits the damage an injected code fragment can perform once executed.
Requiring documented secure development standards and tools enforces use of safe code-generation APIs and escaping.
Mitigating Controls (NIST CSF 2.0) AI
Derived directly from the weakness types (CWEs) cited in the NVD entry via our AI-authored CWE→CSF cross-walk (authority under review) — links open the control.
PR.PS-06's SDLC practices directly target injection flaws via secure coding and testing (mostly), yet as a single broad outcome it leaves many code-generation specifics unaddressed (partial).
PR.DS-10 protects runtime data confidentiality/integrity but has no bearing on neutralizing externally influenced input during code generation, so neither direction shows any preventive effect.
Mitigating Controls (ISO/IEC 27001:2022 Annex A) AI
Derived directly from the weakness types (CWEs) cited in the NVD entry via our AI-authored CWE→ISO cross-walk (authority under review) — links open the control.
Banning unapproved code samples and unauthenticated web services, combined with secure-coding standards and SAST, prevents the dynamic generation or inclusion of attacker-supplied code.
Controls that restrict unauthorized or malicious code from being introduced via external networks or removable media limit opportunities for an attacker to inject and execute arbitrary code.