36 comments

[ 2.7 ms ] story [ 87.7 ms ] thread
A more accurate title would be "Reddit user claims Siri/Cortana transactions are recorded." He doesn't supply any proof.
If someone claimed the sky was blue, would you require them to provide proof before believing them?
Yes. My observations count as a type of proof.
Objection: argumentative.

And yes, if the claim didn't match with my observations I would expect that they have or be prepared to provide proof in some reaonable form. "Extraordinary claims .." and all that.

The point is that the claim does match and is not extraordinary in any way.
This story is old. The user is just confirming it by his experience.
Only "recorded" in the sense that your Google/Bing/Yahoo searches are "recorded," or your browsing history is "recorded". You don't think your Google Now voice searches aren't?

I knew that my searches were probably tracked when I started using Cortana (similar to any other search - unless you're using DDG,) and personally I'm ok with it if it results in a great experience for me. And I'm amazed almost daily at how solid Cortana (and Google voice search for that matter) is.

It is a BIT different to keep a voice recording, because you can recognize a voice recording, unlike a written sentence.
Not to mention the recording of your surroundings, and audio from before and after the query.

Such a recording contains quite a bit more than your average search engine query.

And your search engine query can contain your IP address, geographic information, unique browser fingerprint, and your account information if you are logged in. If Google, Apple, Microsoft, or some malicious entity with access to their data wants to identify you, it is likely trivially easy whether through voice or text data.
Oh, absolutely. The difference is that with voice, they also have the potential to identify the people with you, what you're watching on TV, what radio station is on, etc.
I'm confident that the usual wiretapping laws will be enforced in this area, just as if an ordinary citizen recorded someone's phone call.
When you start using these apps there's generally a popup like this one:

    When you use voice activation commands like
    "OK Google" or touching a microphone icon, a
    recording of the next thing you say, plus a
    few seconds before, may be used and stored
    by Google and associated with your Google
    Account to help recognize your voice and
    improve speech recognition. [1]
Wiretapping is when someone records you secretly, so I don't see how the laws would apply here. It's just like you can record your phone calls if you answer them with "hi, this is SilasX, and this call is being recorded."

[1] http://cdn3.pcadvisor.co.uk/cmsdata/features/3535224/OK_Goog...

I know that consent would make it okay :-P It's just that:

a) It doesn't alert you each time, b) you can use the Siri without getting such a notification (at least not frequently enough that I remember receiving one), c) Siri doesn't answer with "this is Siri and this call is being recorded", d) it's easy to trigger the recorder accidentally, e) it's not even clear to the user that the request is sent home for processing (rather than being handled locally), and f) this wouldn't cover bystanders.

a-b) Affirmative consent isn't required every time...Once you've agreed -- which you're required to do the first time you use Siri/Cortona/Now -- they've fulfilled the legal requirement.

c) Why would it? You've already consented.

d) So?

e) It tells you that your voice data may be processed online, it requires a data connection for processing.

f) There's no law on the books anywhere, as far as I'm aware, that covers bystanders.

for f), There are lots of states where it's illegal to record someone in real life without their consent or their being aware they're recorded. If someone used Siri next to me while I was talking, they'd (or maybe Apple) be breaking the law in some states.
That's only if you have intent. Accidental or incidental recording is not illegal in any of those states. Otherwise every person making a phone call in public could possibly be in violation.
I was under the impression that this was common knowledge. Here's a 2-year-old article about it in Wired: http://www.wired.com/2013/04/siri-two-years/

---

Once the voice recording is six months old, Apple “disassociates” your user number from the clip, deleting the number from the voice file. But it keeps these disassociated files for up to 18 more months for testing and product improvement purposes.

“Apple may keep anonymized Siri data for up to two years,” Muller says “If a user turns Siri off, both identifiers are deleted immediately along with any associated data.”

---

> “Apple may keep anonymized Siri data for up to two years,” Muller says “If a user turns Siri off, both identifiers are deleted immediately along with any associated data.”

Would these mean turning off Siri every week/month and then turning it back on would prevent long term storage of your voice data?

That's assuming they actually delete the data though.
If Apple "disassociates" the data, how can they delete the files when a user turns Siri off? Doesn't that imply they still know which piece of data belonged to a certain user?
(comment deleted)
1) Apple deletes associated data when the user turns Siri off.

2) Apple dis-associates data after six months.

1 & 2 = data older than six months is not deleted, bit is not directly associated with you. So that Siri search for that topic only of interest to communist homosexual terrorist Jews will not have your user ID but might still be identifiable from voice analysis.

You can't escape Macarthyism that easily.

No; they don't disassociate the time it was recorded, only the phone number.
I think the main thing from this claim is that the audio is shared with third parties.
Apple Siri uses Nuance speech recognition technology; afaik Siri run on their server at launch.

It seems they outsourced tasks to another company to improve the speech recognition. The classification task that the Reddit user does can be used by Nuance to improve their data.

As another Reddit user mentioned that after doing the job for a while you will start recognizing speakers: http://www.reddit.com/r/technology/comments/2wzmmr/everythin...

There are hundreds of millions of iOS and Android devices out there. If someone is starting to recognize voices, it supports a common sense reaction that a subset of voice commands are being used for QA/QC purposes to improve accuracy.
That's definitely the case. Plus there are some heavy users while many users try out Siri/GoogleNow/Cortana just a few times and never come back.
Uhm, yes, duh. Isn't this common knowledge? How do people think speech recognition systems are trained? What's more upsetting is that this person is so uninformed about speech systems that they think this is weird, AND they have access. The only people who should have access are the people who are actually doing science.

Ahhh. Reading clarified it: this is someone who got hired to do transcription.

"I'm given an audio file (sound bite) and the corresponding text based translation (how the phone translated the speech). My job is to listen to the file, compare it to the text and provide feed back on how correctly the sound bite was interpreted by the phone. If the text and speech are a perfect match, I just move on. However, if the phone either translated something incorrectly due to a heavy accent or loud background noise, I note that in my evaluation. "

Well, yeah, it might be common knowledge for some, but many don't know, don't care and I enjoy every time it's "revealed" by some community over and over again.

But seriously though, is there a way to do something with Android phone (like using third-party firmware, removing some programs, configuring firewall) to be somewhat sure it's "safe" while still being usable?