Google Assistant voice recordings reviewed by humans include private conversations
Some voice snippets uploaded by Google Assistant are being reviewed by human contractors, as with Amazon Alexa, and can potentially include sensitive conversations, reports say.
Belgium's public broadcaster, VRT, recently gained access to over 1,000 audio files from a Google contractor tasked with reviewing Assistant audio, according to Wired. The audio was captured from devices such as phones, smartspeakers, and security cameras.
Google has acknowledged anonymously transcribing 0.2% of uploads to improve its technology, but instances in which Assistant is triggered accidentally appear to be capturing private information. In the case of a couple in Waasmunster, Belgium, audio included their address, their status as grandparents, and the voices of their son and grandchild.
Google's practices may violate the European Union's General Data Protection Regulations, scholars suggested. One worker screening audio said he encountered a recording in which it sounded like a woman was being physically attacked, but that Google didn't have clear guidelines on what to do in such cases.
A Google spokesperson said that the company has launched an investigation into the contractor mentioned by VRT, on the basis that it violated data security policies.
While Amazon and Google have been subject to the most scrutiny for their voice assistant policies, a technology policy researcher at the Alan Turing Institute in London -- Michael Veale -- has filed an Irish complaint regarding Apple's Siri, saying it violates the GDPR because users can't access uploaded audio. Both Amazon and Google let users review and delete voice samples.
Apple has responded, according to Veale, by saying that its systems handle data well enough that his recordings don't qualify as personal data.
Belgium's public broadcaster, VRT, recently gained access to over 1,000 audio files from a Google contractor tasked with reviewing Assistant audio, according to Wired. The audio was captured from devices such as phones, smartspeakers, and security cameras.
Google has acknowledged anonymously transcribing 0.2% of uploads to improve its technology, but instances in which Assistant is triggered accidentally appear to be capturing private information. In the case of a couple in Waasmunster, Belgium, audio included their address, their status as grandparents, and the voices of their son and grandchild.
Google's practices may violate the European Union's General Data Protection Regulations, scholars suggested. One worker screening audio said he encountered a recording in which it sounded like a woman was being physically attacked, but that Google didn't have clear guidelines on what to do in such cases.
A Google spokesperson said that the company has launched an investigation into the contractor mentioned by VRT, on the basis that it violated data security policies.
While Amazon and Google have been subject to the most scrutiny for their voice assistant policies, a technology policy researcher at the Alan Turing Institute in London -- Michael Veale -- has filed an Irish complaint regarding Apple's Siri, saying it violates the GDPR because users can't access uploaded audio. Both Amazon and Google let users review and delete voice samples.
Apple has responded, according to Veale, by saying that its systems handle data well enough that his recordings don't qualify as personal data.
Comments
Anonymous may not mean you cannot be identified given the time and resources. I'd also be shocked if humans were not involved in understanding problematic Siri recordings as well, and yes those are also anonymized and not connected to specific accounts. If "machines" misunderstood what was said or misinterpreted a wake word then using another machine to understand why it happened hardly seems appropriate.
This statement can be interpreted in many ways:
But one interpretation feels like Google murdered someone with a knife and then handed the subcontractor the bloody knife and then accusing them of murder.
Another interpretation is that Google is investigating a whistle blower, as violating some term in an agreement to not share what Google provided to them.
I'm shocked android users still make excuses for Google.
And no, I'm not listening to your conversations. I just know you have an iKnockoff by the pathetic excuses you provide for Daddy Google.
The problem is, Google alternatives SUCK.
There's a Youtube backlash right now and people are migrating to BitChute. Besides having a terrible name that site is hideous.
Apple needs to enter these spaces now.
https://www.cnbc.com/2019/07/11/google-admits-leaked-private-voice-conversations.html
Gogle admitted on Thursday that more than 1,000 sound recordings of customer conversations with the Google Assistant were leaked by some of its partners to a Belgian news site.
These conversations are used by companies such as Google and Amazon -- which takes clips from the Amazon Echo -- to improve voice responses from their smart assistants. They are supposed to be kept confidential.
But Belgian news site VRT said on Wednesday that a contractor provided it with samples of these sound samples, which VRT then used to identify some of the people in the clips. It also examined the sorts of conversations that Google collects when people say "OK Google," into a phone or a Google Home product. Among other things, VRT heard customer addresses. Sources who talked to the publication also described hearing recordings of a woman in distress and people talking about medical conditions.
That was all in there including blue links to the source report (which wasn't CNBC)
One of the concerns I have about Google is that it appears to be decidedly amoral - it treats data as data and seeks understanding without applying judgement. In a lot of situations, this is fine. But as this article and the source show, the real world contains a lot of ambiguity and judgement needs to be applied so that our society continues to function. The most disturbing line to me was:
"One worker screening audio said he encountered a recording in which it sounded like a woman was being physically attacked, but that Google didn't have clear guidelines on what to do in such cases."
This is a sign of dysfunction. Firstly, there's a conflict experienced by the worker: he knows that sharing this personal information is contrary to the contractual obligation with Google, but also knows that there is a moral obligation to help someone in serious trouble. It's tempting to place blame on the worker, but we don't know how certain he was about what had occurred, how afraid he was of affecting his employment, and a host of other factors. He made the judgement call that it was somebody else's decision but he didn't know how to escalate it.
So, why does Google not have a procedure for dealing with situations like this? Every adult on the planet knows that this sort of thing occurs (as much as we wish it wouldn't), how does the management at Google treat this as a low priority issue or refuse to acknowledge that it is relevant?
On the one hand, we see Google argue that their mission is merely to organise the world's information: OK, fine, if that's the activity being undertaken then that clearly means there is no obligation to act, regardless of the information being managed. But on the other hand, Google is using the analysis of the data to provide services - that too is fine, taken in separation, but it clearly negates the neutrality implied by the stated mission. If you want to collect and analyse ever more intrusive amounts of information and then act on it, you need to act in a manner that benefits society as a whole - do that and you've earned the right to make a profit. If not, then you're a surveillance mechanism that will be resisted to the best of our ability.
<end rant>
EDIT: Just read Apple's white paper on the subject and yes they too acknowledge having humans transcribing what they hear in a users recording compared to what Siri thought was said. I would seriously doubt Apple is sharing what they hear with authorities even if a gunshot followed by a scream is heard. Anonymity and user privacy is of utmost importance.
Also, the issues around if Google has a responsibility to act on what the contractor believes they heard are very complex and we really don't have enough information to know for sure what the right/moral thing would be in this case or if Google already has appropriate policies in place or not. I'm not arguing that.
I'm just saying your weak excuse that if there was a crime being overheard it was in the past and therefore didn't matter anymore is deplorable. If the woman was being attacked as the contractor suspected then how do we know it wasn't an ongoing situation? Even if it wasn't an ongoing situation we don't give up on prosecting crimes because "that was in the past". All crimes were committed in the past. Are you suggesting that unless Google has some sort of Minority Report level of technology then it's irrelevant because the crime has already occurred? That is just ridiculous. There are many factors as to why it may not be possible/appropriate for Google to act on what is overheard by these contractors but you really missed the mark on this one.
I think it's important to consider all sides of an argument and as this site can tend to be very pro apple your comments do sometimes help to broaden the discussion. Sure you tend to be overly pro Google and make comments that ignore sound reasoning on occasion but then there are plenty that make comments on Apple with the same overly biased attitude towards that company. But this goes beyond a slight bias into absurdity. According to this logic should we just remove the notion of a statute of limitations on crimes because if anything has occurred in the past it's too late to act on it now anyway???
Can we just skip that whole process this time and admit that stating that this was "old stuff" was neither relevant to the conversation or a moral/appropriate argument to make?
You continue to ignore/detract from what you are being called out for which unfortunately is your usually approach. I never said it was deplorable for Google not to send the information to the police as you are now trying to imply I did. In fact I said we don't know what policies Google already have in place around this. I acknowledged right from the start that it's a complex issue and yet you continue to try and argue with me about points I never made while ignoring the comment I did make. Or even worse trying to twist it into something I never said.
I'm not trying to attack Google or compare them to Apple, although you seem to be very pre-occupied with the whole Apple vs Google battle. I'm simply saying that the fact the recorded call occurred in the past and only contained "old stuff" is not a reason for it to be dismissed. You have come up with a lot more intelligent talking points after your initial quick jerk reaction but still will not address what you initially said.
It speaks volumes that you still continue to ignore the one critique I raised at the very beginning.
My apologies if I misunderstood the point you were attempting to make when you wrote:
"One worker screening audio said he encountered a recording in which it sounded like a woman was being physically attacked, but that Google didn't have clear guidelines on what to do in such cases."
This is a sign of dysfunction. Firstly, there's a conflict experienced by the worker: he knows that sharing this personal information is contrary to the contractual obligation with Google, but (he) also knows that there is a moral obligation to help someone in serious trouble."
I read that as a concern on your part that an active crime might be underway and no one was helping. My response to you was that it was not a current conversation being transcribed but one from the past, saying and I quote:
"Whatever was heard in those short snippets they were tasked with transcribing was almost certainly old stuff and not actionable for preventing or interfering with whatever was heard."
What wasn't accurate based on what I understood you to be saying? There was nothing that could be done at that time to prevent or interfere with the possible crime taking place. Now if you want to question whether Google (or Apple or Amazon) should report the old snippet to the police for investigation that's an entirely separate conversation.
There is a difference in the amount of identifiable information supplied to those human transcribers between Amazon and Google/Apple according to reports. The former attached the customers first name, whether male or female, account number, and device serial number to the transcribers screen. Both Apple and Google use only anonymized numbered snippets for the human transcription process.