When Helpful Context Leaks: Privacy Risks in Domain-Adapted ASR

Excerpt

arXiv:2605.28211v2 Announce Type: replace Abstract: SpeechLLMs are increasingly deployed in professional settings where domain customisation is standard practice: users supply context in prompts with sensitive information, fine-tune on proprietary recordings, or both. We identify and systematically investigate an overlooked privacy risk of such customisation: a model adapted to recognise domain-specific terminology can be nudged into transcribing a phonetically similar word from its context or t