Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!

 



Forgot your password?
typodupeerror
×
Google Privacy Software Technology

Google Contractors Are Secretly Listening To Your Assistant Recordings (thenextweb.com) 129

A new report from Belgian broadcaster VRT News describes the process by which Google Home recordings end up being listened to by contractors -- and the scary part is that it apparently doesn't take much, if anything, to start a recording. While the recordings are not listened to live, audio clips are sent to subcontractors. The Next Web reports: VRT, with the help of a whistleblower, was able to listen to some of these clips and subsequently heard enough to discern the addresses of several Dutch and Belgian people using Google Home -- in spite of the fact some hadn't even uttered the words "Hey Google," which are supposed to be the device's listening trigger. The person who leaked the recordings was working as a subcontractor to Google, transcribing the audio files for subsequent use in improving its speech recognition. They got in touch with VRT after reading about Amazon Alexa keeping recordings indefinitely.

According to the whistleblower, the recordings presented to them are meant to be carefully annotated, with notes included about the speakers presumed identity and age. From the sound of the report, these transcribers have heard just about everything. Personal information? Bedroom activities? Domestic violence? Yes, yes, and yes. While VRT only listened to recordings from Dutch and Belgian users, the platform the whistleblower showed them had recordings from all over the world -- which means there are probably thousands of other contractors listening to Assistant recordings. The VRT report states that the Google Home Terms of Service don't mention that recordings might be listened to by other humans.
The report says that the recordings are identified by numbers rather than user names, but VRT "was able to pick up enough data from the recordings to find the addresses of the users in question, and even confront some of the users in the recordings -- to their great dismay," reports The Next Web.

What does Google have to say about all this? That they only transcribe and use "about 0.2% of all audio clips" to improve their voice recognition technology.
This discussion has been archived. No new comments can be posted.

Google Contractors Are Secretly Listening To Your Assistant Recordings

Comments Filter:
  • by BytePusher ( 209961 ) on Wednesday July 10, 2019 @09:16PM (#58905416) Homepage
    But the other 99.8% is archived just in case they need a fresh data set or there's a government subpoena.
  • I'm so glad a human might be listening when I am cursing the Google Assistant service. That makes me feel so much better, even if they are just probably laughing at me.

    Frankly the service and it's capability is appalling and does not seem to be improving. It continually leaves me feeling frustrated and dis-empowered. Its not the actual voice recognition which seems to be the problem, rather the contextural application of that data, which they need to improve. Google must know this, yet they seem unable to deliver.

    Three years of machine learning and the machine hasn't learned very much at all. Maybe the humans can help.

    • Re: (Score:2, Insightful)

      by gweihir ( 88907 )

      Contextual application of data needs intelligence. The real thing, not something cheaply faked with some statistical classificators misleadingly labelled machine "learning" and some glitter on top. As we have zero actual intelligence in machines and not even a credible theory how it could be implemented, we are not getting that anytime soon. Incidentally, we have zero clue how human intelligence works either. The human brain seems to be too small and slow for things that smart human beings can do. At the sa

      • Re: (Score:2, Insightful)

        by Etcetera ( 14711 )

        Contextual application of data needs intelligence. The real thing, not something cheaply faked with some statistical classificators misleadingly labelled machine "learning" and some glitter on top. As we have zero actual intelligence in machines and not even a credible theory how it could be implemented, we are not getting that anytime soon. Incidentally, we have zero clue how human intelligence works either. The human brain seems to be too small and slow for things that smart human beings can do. At the same time, by some estimations, the human brain is within one order of magnitude or so the most powerful computer possible in this universe. The whole thing is a complete mystery at this time.

        Why in the world do you think it needs to be intentionally designed or that we'll have a rational mechanism for describing or understand it when it does appear?

        Actual Intelligence in machines will occur spontaneously, probably as a result of hooking a tipping point number of inputs into a tipping point number of outputs. It will be a side-effect, it will be a surprise, and it will be a mystery how it works.

        • Re: (Score:3, Insightful)

          by gweihir ( 88907 )

          Actual Intelligence in machines will occur spontaneously, probably as a result of hooking a tipping point number of inputs into a tipping point number of outputs. It will be a side-effect, it will be a surprise, and it will be a mystery how it works.

          That is techno-mysticism, not science. Digital computers do not work that way. This cannot happen with them.

          • by Etcetera ( 14711 )

            Actual Intelligence in machines will occur spontaneously, probably as a result of hooking a tipping point number of inputs into a tipping point number of outputs. It will be a side-effect, it will be a surprise, and it will be a mystery how it works.

            That is techno-mysticism, not science. Digital computers do not work that way. This cannot happen with them.

            You're referring to things at a very low level. Imagine a neural net that had as inputs all live data from Facebook and Twitter, all live data from Siri, Alexa, and Google Home, and as outputs controlled Amazon feeds, Google Adwords, and FB/Twitter algorithms and was set to run deep learning simulations on improving the mood of those providing the input. The fact that it's running on "digital computers" doesn't change the complexity at issue.

        • by Kjella ( 173770 )

          Actual Intelligence in machines will occur spontaneously, probably as a result of hooking a tipping point number of inputs into a tipping point number of outputs. It will be a side-effect, it will be a surprise, and it will be a mystery how it works.

          I doubt we could create real intelligence without soon after discovering how it works. A computer doesn't have any undefined behavior or state, even if we didn't understand it at first we could copy it, reset it, log it and run millions of simulations to discover what bits are essential and analyze them in detail, step by step. The brain is an incredible device, but we also know from people with brain damage that huge parts of it are sensory perception, motor control, emotions, memory etc. and that problem

      • Contextual application of data needs intelligence.

        Really? Then do I misunderstand what you mean with "contextual application of data"?

        As shown by voice assistants, text analysis systems etc, I would say that it is obvious that some degree of contextual application of data is indeed possible without intelligence. I agree that for whatever definition of "intelligence" chosen, this isn't it, but the applications work.

        I see it as a scale: Some systems as above can handle some contexts to some degree. Some humans can handle more contexts to greater degrees (and

        • by gweihir ( 88907 )

          If you have a fixed, known context, this is simple. Bit the situation here is, obviously, that the context is variable and needs to be determined. And that needs real intelligence.

    • Maybe the humans can help.

      Unlikely since you haven't given them a shred to go on. Perhaps a specific example or two would help?

      • by cute-boy ( 62961 )

        If you show me you are a Google employed or created entity, I am happy to provide you with examples here. I've given Google plenty of feedback via the "OK Google, send feedback" channel, which as far as I can tell, has not been acted upon by humans or machines.

        • by AmiMoJo ( 196126 )

          Maybe you could just share them with us?

          I use Assistant and Translate quite a bit. Transcription of my voice is near flawless, even with the TV on in the background or in public places. It's also pretty much instant.

          What I find most impressive is that it handles names well. Asking it to play Ecliptica by Sonata Arctica works. Asking it to send messages to people with Japanese and Chinese names works. That's where every other voice assistant I've tried fails.

          Having said that, English is definitely better tha

        • If you show me you are a Google employed or created entity, I am happy to provide you with examples here.

          That's OK -- it turns out I didn't have to look far to find some of your prior examples [slashdot.org] -- they're about as silly and overwrought as I would have expected from the signal/noise ratio of your original post.

          In the future I'll just keep in mind you clearly don't care for Google in general, and let you emote.

    • "when I am cursing the Google Assistant service"

      How often do you suppose someone, somewhere in the world says "fuck you, Google!" to their Android phone or other surveillance endpoint? I'm gonna guess, worldwide, it happens a couple hundred times per minute. Big Brother Google surely knows how often; but just as surely will never tell us.

    • I was thinking the very same thing. At least they have to listen to me say, "Hey Google, you're useless." and "You've failed me once again." and "Hang your head in shame." The only reason I don't say, "Please go kill yourself" is because I'm worried it'll start giving me anti-suicide advice.
  • Why? (Score:5, Insightful)

    by mschaffer ( 97223 ) on Wednesday July 10, 2019 @09:27PM (#58905474)

    Why would you knowingly allow these kind of devices into your home or office?

    • by gweihir ( 88907 )

      People do all kinds of stupid things, this is just one of them.

    • Re: (Score:3, Insightful)

      by Rockoon ( 1252108 )

      Why would you knowingly allow these kind of devices into your home or office?

      Your smart phone is one of those kinds of devices. Do you leave it outside your home or office?

      • My smartphone is powered on and connected all the time. True. However, it lacks the battery power to perform sophisticated analysis or simply transmit everything all the time. For now. And listening in all the time is not its dedicated, intentended purpose.

        I know that YouTube and Amazon present products and topics talked about to in the last 24-48 hours while the phone was in the vicinity, even if no input and no search on that topic had been performed and even if the person talket to has zero connection to

      • Whilst my phone is technically capable, it's clearly not sending my every word to $someone - if it were, we'd see it on my data bill, or indeed my battery budget.

        A home assistant though, is *designed* to have the (omnidirectional) microphone on all the time, to "listen" to that microphone for the activation words and then to communicate with the mother ship. This design is common to all assistants, and all assistants have been shown to activate when no activation words were actually spoken.

        Sure, someone cou

        • Whilst my phone is technically capable, it's clearly not sending my every word to $someone - if it were, we'd see it on my data bill, or indeed my battery budget.

          Sure, because everything gets billed... no corporation has ever made a deal to carry un-metered secret/un-documented data... ever! 'cept for the dozens that we already know about.

          ..and what would you know of the battery life? If its always doing it, then it has always been effecting your battery life. The battery life you think is normal could in fact include regular dumps of very low bitrate audio that is less than 1000 bits per second, which is below 0.01% of 3G (not even 3.5G) bandwidth. Your phone is

    • This is the same generation that sends dickpics even after Snowden told everyone that the NSA collects everything and stores it for future use.
      This generation does not care about privacy.

      • "This generation does care about privacy, but sees no viable way to fight the combined forces of the police state and monopolist big business"

        FTFY

    • Two ways to look at it.

      One is that people don't care about privacy - they are willing to trade all their privacy for a little convenience or a little money. That is the typical response.

      There is another possibility: people have discovered that they *can't* have privacy. You can not have home assistants. Or voice activated TVs (does the disable really work?). Or cell phones. Or one of many modern cars that have microphones (like onstar), or home computers with a wider range of apps with microphone ac

      • There's a third option - people weren't aware there was a privacy issue when they purchased the device.

        My sister bought a bunch of Amazon Echo devices a couple years ago. She started with the tall one, then her kids wanted the smaller ones (if I remember correctly, there are semi-portable ones or small ones which can become semi-portable with a particular attachment). She thought they were great. But, as the stories have come out (which I've made sure she's aware of), she's learned that these are privacy-in

      • by AmiMoJo ( 196126 )

        I read a book co-written by Arthur C. Clarke about a future where surveillance was total and available to everyone. Some tech allowed you to open a wormhole and look through it, and it was undetectable. Privacy ceased to exist.

        Well, almost. Some people started living in the dark, communicating only by tapping on each other's hands (like a chorded keyboard). Maybe that's the future.

        Or we could just do something about it. Make it illegal, require microphones to have a physical off switch etc.

    • Why would you knowingly allow these kind of devices into your home or office?

      Because people don't care? Privacy is not considered the same for everything everywhere. At this point it is almost impossible to consider ignorance as the cause which means people have put active thought into idea the a corporation is recording their conversations.

      As for not uttering the words "Ok Google" that doesn't surprise me. It may be a trigger word but it has an incredible number of false positives. When a parrot can activate and order items online using Alexa, a child with its illformed language as

    • You've got a cellular phone, right? You carry one everywhere you go.

    • by Baleet ( 4705757 )
      So I can play like I'm in the future. Since I'm still waiting on my jet pack and my flying car.
  • Well, of course I do not have one of these listening devices. But I expect that quite a few people did trust that no human ever listens in on their living-room (and apparently now also bedroom).

    • Well, of course I do not have one of these listening devices.

      Remember this is Google Assistant, which is not just on Google Home but also one most (all?) Android phones.

      • by gweihir ( 88907 )

        Most Android phones, not all. And the function has to be activated first and cannot be always on afterwards. Would be a criminal act otherwise, for example in Germany.

  • "[T]he words "Hey Google," [...] are supposed to be the device's listening trigger."
    It's all said there. How can the device NOT be listening all the time ??

    • by Anonymous Coward

      The way these things are sold is that they explain that there's a "low power" process that's constantly checking for the "wake word" and that full processing, including sending audio to Google, only happens after the "wake word" is heard.

      This is a half-truth.

      There is, in fact, a very low-power process that constantly checks the audio input. Its job is to start a secondary process if it detects actual audio on the microphone. This process starts recording and then uses a very low fidelity check to see if it

    • It's amazing how often something else can get mistaken for that wake-up phrase. I have my Android phone hooked up to the car for listening to podcasts. I was surprised at how often the audio mutes and Android Auto goes into listening mode because someone on a podcast has said something which it has mistaken for "hey Google"
  • Child privacy laws they need to come down hard on stuff like this.
    also in the USA we need real laws to help people not EULA's that take away your rights!

    • EULA's cant take your right away, regardless of whats in them.

      At best they are a contract (but they probably arent), and the remedies for breach of contract happen in a court of law, which fully recognizes rights regardless of whats in the contract, and will "make whole" the wronged party not by stripping rights but instead by awarding money.

      The fact is that these things rarely get challenged in court because the sums of money are always tiny in comparison to the cost of lawyers. At best they are dealt
      • EULAs impair both rights and remedies: Arbitration clauses allow abusive corporations to keep their behavior and losses secret; No Class Action clauses make it uneconomical to pursue many claims. Clauses like these are frequently unfair, but unless you were tricked into agreeing (or the terms shock the conscience of a judge who deals with unfair contracts all the time), your rights will be as described in the contract.

        (Also, the worst trouble I've heard with small claims collection is that it is easier for

        • The corporation is free to argue that the EULA requires that the user go through an arbiter...

          ....an argument that they will have to make in court, in front of a judge.

          Just because the EULA says it doesnt make it so, and you cant point to the EULA as the argument that does make it so. Its circular logic. Stop it.
    • No, everybody needs privacy rights, not just children.
  • by Ol Olsoc ( 1175323 ) on Wednesday July 10, 2019 @10:24PM (#58905684)
    Husband: "I'm kind of concerned, dear".

    Wife: "About what, honey?"

    Husband:Oh, just thiinking that Google Assistant might be listening in on everything we say".

    Wife: "Can it really do that?

    Husband: Oh, probably not. (laughs)

    Wife: (laughs)

    Google Assistant: (laughs)

  • If it can place sound on the internet - it will be used to record whatever Google, or law enforcement wants to record with it.

    Deal with it. If you are okay with being 24/7 monitored, buy one. If not, you probably shouldn't

  • As in the topic, is it all the same, as no "Hey Google" needed?
    • As in the topic, is it all the same, as no "Hey Google" needed?

      Yes, it is all the same. Fortunately you can remove the right to use the microphone from the app. Android complains loudly when you do though, but it does remove the right.

  • bring a big ad brand microphone into your home.
  • devils advocate, but wouldn't the activation phrase, by nature, not appear on the recording? if it did, that would mean that google assistant WAS really listening all the time, and just not saving the audio.

    It strikes me now, that this is probably exactly how they are supposed to work. Always on, always listening.

    • by HiThere ( 15173 )

      They have to always be listening, or they couldn't hear and recognize the activation phrase.

  • Surprised? LOL (Score:5, Insightful)

    by JustAnotherOldGuy ( 4145623 ) on Wednesday July 10, 2019 @11:51PM (#58905920) Journal

    OF COURSE they're doing this. Anyone who is surprised by this either hasn't been paying attention or has just awoken from a 20-year coma.

    I knew something like this would be discovered as soon as I saw these "always listening" devices start appearing.

    Parse the phrase "always listening"...try and puzzle out what the meaning of that mysterious, hard-to-understand phrase is.

    • There's always listening, and then there's always listening for activation words. Unfortunately by necessity the activation phrase is treated very loosely causing recordings to often start at times the phrase wasn't said.

      There certainly is no "always listening". It's not recording eternally and continuously recording your conversation and beaming it to the G-borg.

      So while you were being facetious you actually hit the nail on the head. The impact of "always listening" is actually quite hard to define and tru

      • So while you were being facetious you actually hit the nail on the head. The impact of "always listening" is actually quite hard to define and truly understand.

        Also, it's worth noting that you pretty much always know when the device has triggered, because it always says something, even if it's just "I didn't understand". Anyone who has one knows that they do occasionally trigger when no one has said the hotword.

        Also, with Google's devices, at least, you can look at your Google Home account and see a list of all of the trigger events, and listen to whatever was recorded. And delete the recordings. (And, yes, they will be deleted, as in permanently and entirely

        • "Also, with Google's devices, at least, you can look at your Google Home account and see a list of all of the trigger events, and listen to whatever was recorded. And delete the recordings. (And, yes, they will be deleted, as in permanently and entirely gone; no archives kept.)"

          That's a remarkable display of trust in a public corporation, which is to say, insanity.

          Believing that Google deletes "your" recordings from their servers is nonsensical, let alone believing that they eliminate all backup copies.

          • "Also, with Google's devices, at least, you can look at your Google Home account and see a list of all of the trigger events, and listen to whatever was recorded. And delete the recordings. (And, yes, they will be deleted, as in permanently and entirely gone; no archives kept.)"

            That's a remarkable display of trust in a public corporation, which is to say, insanity.

            Believing that Google deletes "your" recordings from their servers is nonsensical, let alone believing that they eliminate all backup copies.

            It's not the slightest bit insane for me, since I know the people who build the relevant systems. And, while it would be moderately painful to erase relevant sections of the backup tapes, it's easy to destroy the encryption keys. So there's no need to track down backup tapes.

            Also, note that if Google says that something is deleted, but lies, I believe that would put them in violation of the FTC consent decree, and at risk of having that violation discovered in an audit. That would be bad, and not someth

            • "rational people who don't have my personal connections can also have a high degree of confidence that Google tells the truth about such things, because of government regulators."

              Which government do you believe will regulate them in good faith?

              • "rational people who don't have my personal connections can also have a high degree of confidence that Google tells the truth about such things, because of government regulators."

                Which government do you believe will regulate them in good faith?

                I already said that: The FTC, which, as you should know, is an agency of the US government.

                • I already said that: The FTC, which, as you should know, is an agency of the US government.

                  Many suspect that Google is effectively an agency of the US government, like Microsoft. Guess we'll find out.

        • Also, it's worth noting that you pretty much always know when the device has triggered, because it always says something

          You think it always says something, but does it really? How would you know if it opted not to say something?

          Also, with Google's devices, at least, you can look at your Google Home account and see a list of all of the trigger events, and listen to whatever was recorded. And delete the recordings. (And, yes, they will be deleted, as in permanently and entirely gone; no archives kept.)

          1) You'll only see what they let you see.

          2) Yes, they'll be *cough* deleted forever, trust us, Google would never do anything naughty or fib to you! They'll totally for sure be deleted forever!

          I'm just playing Devil's Advocate here, but at the same time, you have to admit that we don't really know what this gadget is doing.

  • The 0.2% sounds ok perhaps, but given google's account size it's still potentially millions of recordings. They also don't state the timeframe of the percentage. Is it per day? Since they say "recordings" and not "accounts", means everyone may have their recordings listened to, eventually.
  • by tinkerton ( 199273 ) on Thursday July 11, 2019 @04:13AM (#58906530)

    There is a microphone button in my browser (for google search)which i inadvertently switch on very often . That is a lot of recordings of background noise which google has no business with

  • _all_ 'smart'speakers very much likely have the same operations model behind them, and this is not just a google problem.

  • Comment removed based on user account deletion
  • Or, get a flatulent dog and keep the thing in the kennel when you're not actually using it. That's if you can find a use for it, which it certainly hasn't advertised any for here.

He has not acquired a fortune; the fortune has acquired him. -- Bion

Working...