speech-to-text

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). In the ElevenLabs Speech-to-Text runtime workflow, the LLM ingests transcription text derived directly from user-supplied audio streamed or uploaded via the required API/WebSocket inputs (e.g., client.speech_to_text.convert(file=audio_file, ...) or input_audio_chunk/real-time audio_base_64), meaning outsider-authored free text can be supplied inside that audio payload.