Back to the MIT repository
2. Privacy & Security3 - Other

Hardware Vulnerabilities

The vulnerabilities of hardware systems for training and inferencing brings issues to LLM-based applications.

Source: MIT AI Risk Repositorymit24

ENTITY

3 - Other

INTENT

2 - Unintentional

TIMING

3 - Other

Risk ID

mit24

Domain lineage

2. Privacy & Security

186 mapped risks

2.2 > AI system security vulnerabilities and attacks

Mitigation strategy

1. Implement a comprehensive hardware supply chain security program, including rigorous vetting of suppliers and the integration of secure design practices during the component development lifecycle, to prevent the introduction of fundamental weaknesses. 2. Utilize advanced, data-driven methodologies, such as LLM-assisted hybrid platforms, for the zero-shot discovery, classification, and systematic assessment of hardware-related CVEs to proactively identify systemic risks. 3. Establish protocols for routine patching and firmware updates across all deployed hardware components to mitigate known vulnerabilities promptly and ensure ongoing system resilience.