Researchers Demonstrate Inaudible Audio Attacks That Hijack AI Voice Assistants

Thumbnail Image

The information displayed in the AIM should not be reported as representing the official views of the OECD or of its member countries.

Researchers from China and Singapore have demonstrated that inaudible, adversarial audio signals embedded in podcasts, videos, or music can hijack AI voice assistants and smart speakers. These attacks can trigger unauthorized actions, such as accessing sensitive data or sending information to attackers, highlighting serious security vulnerabilities in AI-powered audio systems.[AI generated]

Why's our monitor labelling this an incident or hazard?

The event involves AI systems explicitly—voice AI models used in smart speakers and phones. The hackers' adversarial audio exploits the AI system's malfunction or vulnerability to cause unauthorized actions, directly leading to harm in terms of privacy breaches and potential access to sensitive information. The harm is realized or highly plausible given the demonstrated attack method. Therefore, this qualifies as an AI Incident due to direct harm caused by the AI system's exploitation.[AI generated]
AI principles
Privacy & data governanceRobustness & digital security

Industries
Consumer productsDigital security

Affected stakeholders
Consumers

Harm types
Human or fundamental rights

AI system task:
Interaction support/chatbots


Articles about this incident or hazard