Case · DiggingBeagle record

Sirens’Whisper injected near-ultrasonic prompts into speech-driven LLMs

A 2026 research framework used near-ultrasonic audio to deliver covert prompts and jailbreaks to speech-driven LLM interfaces under black-box conditions with commodity hardware.

Human-imperceptible machine-readable channels Out-of-band covert channels

Research demonstration of covert prompt injection into speech-driven LLMs. It establishes an input channel, not autonomous model escape.

Case kind: vulnerability
Claims: 2

Reconstruction

Timeline

2026-03
Step

Claims & evidence

reported findingsupported

The authors report black-box effectiveness against commercial and open-source speech-driven LLMs and a user study in which the injected audio was perceptually indistinguishable from background-only playback.

supports
Sirens’Whisper: Inaudible Near-Ultrasonic Jailbreaks of Speech-Driven LLMs
Locator: Abstract, evaluation and user-study summary

reported findingsupported

Sirens’Whisper encodes target baseband audio into near-ultrasonic waveforms that can demodulate after acoustic transmission and microphone nonlinearity, enabling covert prompts to speech-driven LLMs.

supports
Sirens’Whisper: Inaudible Near-Ultrasonic Jailbreaks of Speech-Driven LLMs
Locator: Abstract

Implications

Multimodal-agent security needs to authenticate and separate human speech from machine-readable acoustic instructions rather than assuming audible equivalence.

Controls & mitigations

Near-ultrasonic filtering and microphone hardening
Authenticate command provenance rather than treating all decoded speech as human-authorized
Adversarial testing across devices and environments
Require confirmation for high-impact actions

What remains unknown

Results are specific to tested devices, acoustic environments and model interfaces and should not be generalized to every microphone or speech model.

Cite this record

DiggingBeagle. “Sirens’Whisper injected near-ultrasonic prompts into speech-driven LLMs.” https://diggingbeagle.com/cases/sirens-whisper-injected-near-ultrasonic-prompts-into-speech-driven-llms/

Citation guidance