Apple's Siri Voice Clip Storage Policy Raises Privacy Concerns

NEWS ANALYSIS: For the first six months, the files are associated through a random number tied to your copy of Siri, but after that, the number is stripped.

Apple’s revelation to Wired that it stores for two years the voice clips for the questions that Siri users ask the automated assistant at least clarifies some details about Siri that Apple hasn’t revealed previously.

While Apple has made it clear that Siri actually works by digitizing your voice when you ask Siri a question and then sending that voice clip to Apple’s servers for analysis, it’s never really said what it does with the clips after you have completed your inquiry.

Siri works like this. When you press the home button on your Siri-enabled device, all the Siri app does is digitize your question, and then send it to Apple. There the Apple servers use voice recognition software to parse out what you said, and then sees if it has the answer available in its databases. If you ask Siri, “What’s the best restaurant within five miles,” Siri will send your location along with the question.

Once the question gets to the Apple servers, the servers will look for all of the restaurants within five miles of your location and the ratings using an online database, usually Yelp. Once it looks up that information it will send you a list sorted by star rating of the restaurants within that distance.

While Siri knows about restaurants, bars and sports data, there’s a lot it doesn’t know about. So when I ask for the current score in the Washington Nationals game, Siri knows the answer. When I ask Siri how long Apple keeps my voice clips, it asks if I want it to search the Web. If I say I do, then Siri goes off to wherever questions go and never comes back with an answer.

What happens then is that Apple keeps the voice clip and analyzes it to see how to make Siri more likely to find the answer. In testing I’ve found that Siri rarely knows any answer outside of the specific areas pre-programmed by Apple. This means that except for knowing the score for the Nationals on a given night, most of what I’ve accomplished is to give Apple fodder for improving the search process. I guess that’s nice, but not particularly helpful.

For example, I asked Siri how long Apple stores my voice clips. Siri didn’t know, and asked if I wanted to search the Web. I told Siri to search, but the assistant never revealed the answer.

Then I tried one of the example scenarios that have appeared in the technology press—questions that you shouldn’t ask Siri because someone might find out. “Siri,” I asked, “Where can I go to get drunk?” This, it seems, is one question that has a pre-programmed response. “I hope you’re not driving, Wayne,” it said. Then Siri presented me with a button labeled “Call a taxi.”

Wayne Rash

Wayne Rash

Wayne Rash is a freelance writer and editor with a 35 year history covering technology. He’s a frequent speaker on business, technology issues and enterprise computing. He covers Washington and...