AI assistants can be hijacked and manipulated by inaudible sounds
Hidden audio commands can hijack AI voice assistants and transcription tools without users hearing anything unusual, according to new research set to be presented at the IEEE Symposium on Security and Privacy next week. The study shows that carefully crafted audio clips can elicit unauthorized actions from audio-language models (LALMs), including downloading files, sending emails, … The post AI assistants can be hijacked and manipulated by inaudible sounds appeared first on CyberInsider.
AI assistants can be hijacked and manipulated by inaudible sounds through 'AudioHijack', a form of auditory prompt injection that embeds malicious instructions in ordinary audio, allowing attackers to trigger unauthorized actions such as downloading files or sending emails.
Read the full article
This is a curated summary. The complete article is available at Cyber Insider.