Privacy groups have been asking after Siri's info hoarding habits for quite a while, but this is the first time Apple has given out the explicit details. Although Apple does keep your data, it says the snippets it saves are completely anonymous and only used to improve Siri.
Here's how it works: when you talk to Siri, your conversation is sent to Apple for analysis. Random numbers are generated to represent you and the voice data associated with the virtual assistant. This code is different from both your Apple user ID and your UDID, but it's filed with your Siri logs. Once a specific file is six months old, Wired says the number is "dissociated" from the clip, but the clip is still kept on hand for up to 18 months. And if you turn Siri off, all of the identifying factors are deleted. So even though it's all anonymous, perhaps you'll think twice about what you're gabbing with Siri about next time. [Wired]
Apple Keeps Your Siri Data for Two Years