CVE Allocation: Why AI Models Should Be Excluded

NVIDIA contends that the Common Vulnerabilities and Exposures (CVE) system, a cornerstone of software security, should exclude AI models. The CVE framework, managed by MITRE and backed by CISA, traditionally identifies flaws in software applications and frameworks. Extending it to AI models introduces complexities, as their failure modes—adversarial prompts, data poisoning—don’t align with CVE’s Core focus on breaches of confidentiality, integrity, or availability.

AI-related vulnerabilities often stem from the frameworks or applications leveraging the models, not the models themselves. NVIDIA’s stance underscores a growing debate: as AI integrates deeper into enterprise systems, security protocols must evolve—but perhaps not by conflating model-specific risks with traditional software vulnerabilities.