APromptRiskDBThreat intelligence atlas
AI Risk

Vulnerabilities arising from additional modalities in multimodal models

"Additional modalities can introduce new attack vectors in multimodal models as well as expand the scope of the previous attacks, ranging from jailbreaking to poisoning [13]. Typically, different modalities have different robustness levels, allowing malicious actors to choose the most vulnerable part of the model to attack [119, 181]."

AI Risk2. Privacy & Security2.2 > AI system security vulnerabilities and attacks3 - Other

Record summary

A quick snapshot of what this page covers.

Techniques29Attack methods connected to this risk.
Mitigations25Defenses that may help with related attacks.
Domain2. Privacy & SecurityThe broad risk area this belongs to.

Risk profile

How this risk is described and categorized.

Domain2. Privacy & Security
Subdomain2.2 > AI system security vulnerabilities and attacks
Entity3 - Other
Intent3 - Other
Timing3 - Other
CategoryAttacks on GPAIs/GPAI Failure Modes
SubcategoryVulnerabilities arising from additional modalities in multimodal models

Suggested mitigations

Defenses that may help with related attacks.

Validate AI Model

ML Model EvaluationMonitoring and Maintenance
LifecycleML Model Evaluation + 1 moreCategoryTechnical - ML

Code Signing

Deployment
LifecycleDeploymentCategoryTechnical - Cyber

Generative AI Guardrails

ML Model EngineeringML Model Evaluation+1 more
LifecycleML Model Engineering + 2 moreCategoryTechnical - ML

Generative AI Guidelines

ML Model EngineeringML Model Evaluation+1 more
LifecycleML Model Engineering + 2 moreCategoryTechnical - ML

Sanitize Training Data

Business and Data UnderstandingData Preparation+1 more
LifecycleBusiness and Data Understanding + 2 moreCategoryTechnical - ML

AI Bill of Materials

Business and Data UnderstandingData Preparation+1 more
LifecycleBusiness and Data Understanding + 2 moreCategoryPolicy

AI Telemetry Logging

DeploymentMonitoring and Maintenance
LifecycleDeployment + 1 moreCategoryTechnical - Cyber

Memory Hardening

ML Model EngineeringDeployment+1 more
LifecycleML Model Engineering + 2 moreCategoryTechnical - ML

Verify AI Artifacts

Business and Data UnderstandingData Preparation+1 more
LifecycleBusiness and Data Understanding + 2 moreCategoryTechnical - Cyber

Model Hardening

Data PreparationML Model Engineering
LifecycleData Preparation + 1 moreCategoryTechnical - ML

Input Restoration

Data PreparationML Model Evaluation+2 more
LifecycleData Preparation + 3 moreCategoryTechnical - ML

Source

Research source for this risk, when available.