Biometric Notice and Consent FAQs

Q: What should I know about biometric laws?

A: Biometric laws are privacy law focused on the handling of Biometric Data (as defined in the Biometric Data Addendum), such as a voiceprint or speaker embedding, and certain related information. They may apply if a service you are using involves Biometric Data. Many biometrics laws have specific requirements for providing notice to and obtaining consent from individuals who the Biometric Data relates to and complying with deletion requests. AssemblyAI’s customers are responsible for understanding and complying with their obligations under biometric laws, including by providing required notices and obtaining required consents. If you provide notice and obtain consent, you should do so on behalf of any relevant service providers, including AssemblyAI (identified as your service provider). Please see our Terms of Service and Biometric Data Addendum for more information about our biometric features. You can find a sample notice and consent language for service providers below.

Q: Does AssemblyAI have sample language for providing notice and consent on behalf of service providers?

A: Yes. Below is sample language for providing notice and consent on behalf of service providers (such as AssemblyAI) in connection with Biometric Data (as defined in the Biometric Data Addendum, available at: https://www.assemblyai.com/legal/biometric-data-addendum). You are responsible for making your own assessment of whether your use of AssemblyAI’s Services (as defined in the Terms of Service or other written or digital agreement between you and AssemblyAI (“Agreement”) meets applicable legal requirements. Please note that this sample language is not legal advice and AssemblyAI makes no representations, warranties, or covenants as to the appropriateness, sufficiency or accuracy of this sample language, nor that this sample language will comply with all applicable laws, including but not limited to Biometric Data Laws (as defined in the Biometric Data Addendum).

[Your company name] (“Company”) uses one or more service provider(s) for [describe the purpose, e.g., “speech-to-text services”]. Biometric identifiers and biometric information, which include [describe the type of data at issue, e.g. voice prints or speaker embeddings] (collectively, “Biometric Data”), may be collected, stored, and used by Company’s service provider(s) on behalf of Company for the purpose of providing the service. Company will instruct its service provider(s) to permanently destroy Biometric Data that is stored on Company’s behalf when the initial purpose for collecting or obtaining such data has been satisfied, when you request that the data be deleted, or earlier, if required by law. Biometric Data may be transmitted between Company and its service provider(s) to provide and receive this service. By [executing/agreeing to] this consent, you hereby agree to allow Company and its service provider(s) to collect, use, and store your biometric data as described above.

Q: Does AssemblyAI retain speaker diarization speaker embeddings?

A: Our current speaker diarization feature (which only applies if you opt in), involves the detection of distinct speakers in an audio recording that you provide for transcription (e.g., Speaker A, Speaker B) and as such, related speaker embeddings may be deemed biometric data depending on the applicable privacy law. Speaker embeddings are created on a recording-by-recording basis and are not retained after the transcription of the recording is provided to you. We have no way of identifying the actual identity of the speaker.

Q: How long does AssemblyAI retain voice clone reference audio files and the resulting voice clones?

A: You may delete a voice clone at any time in your AssemblyAI dashboard or by making an API call (see Documentation). The voice clone reference audio file that was used to generate the voice clone is deleted at the same time.

AssemblyAI automatically deletes voice clone reference audio clips and the resulting voice clone sixty (60) days after the voice clone was created or consent was extended. You may instruct AssemblyAI to refresh each 60 day retention period for each voice clone, and its voice clone reference audio file, by changing the API parameter of “extend_biometric_consent” from false to true (the parameter is set to false by default). If you set this parameter to true, this is your instruction to extend the retention period of each voice clone, and its voice reference audio file by sixty (60) additional days.