Apple devices may be sharing your audio recordings and transcripts with the company through the Improve Siri & Dictation setting. When enabled, this toggle allows Apple to store and review the transcripts and audio of your interactions with Siri, as well as translation or dictation requests. Users are asked to enable this during device setup, but may have done so without fully understanding Apple's data processing and retention policies. The speech processing models supported by this data are essential to Apple's AI initiatives, particularly with the new AI-powered Siri in iOS 27 and advanced dictation features.

Apple's Improve Siri & Dictation setting allows employees to review audio, transcripts, and related request data of your voice command requests. This includes when and where you made the request, your device's specifications, performance statistics, and the category of interaction. Apple may also review contextual information such as your contact list or downloaded apps. According to Apple's legal page, employees review these transcripts to fine-tune Siri, Search, Voice Control, Translate, and automatic speech recognition models. Only a select group of interactions is chosen for review.

Typically, Apple retains voice data for up to two years, though some data may remain beyond this window. The information is associated with a random, device-generated identifier that rotates multiple times per hour, rather than being tied directly to your Apple Account. Apple states that your request history is not sold to marketers, and access is limited to employees subject to strict confidentiality obligations. However, a subset may be reviewed by graders, and trusted third-party service providers may also process and store this information.