Google spying is multilingual

Last week, we learned how Google helped Hillary Clinton sway anywhere between 2.6 and 10 million votes. Google’s parent company was Clinton’s strongest financial supporter in 2016. Since his public testimony before a Senate Committee, Dr. Robert Epstein‘s report titled “A Method for Detecting Bias in Search Rankings, with Evidence of Systematic Bias Related to the 2016 Presidential Election” where he explained how Google used search manipulation, suggestion manipulation in search results and subliminal messaging so Google users ended up deciding to vote, has been removed from Scribd.com In the last few hours, we have learned that the tech giant is not only spying, but that its spying is multilingual. In explaining what companies like Google and Facebook are doing behind the scenes, Dr. Robert Epstein warned that these companies represent a threat to Democracy. Here is his complete testimony: Google listens to private conversations in Spanish and Portuguese. The company admitted just a week ago that “language experts” listen to 0.2% of the conversations that users keep with their virtual assistant after a leak of audio in Dutch language to a Belgian television. Google admitted to the spying although it did not specify in which languages ​​his so-called analysts listen. We also learned that Google’s spying is not new. They have been doing this for years with private recordings in different languages, including Spanish and Portuguese. Listening to audios is done, in principle, to improve the functioning of Google Assistant. That is, for the assistant to be able to better understand what the users are talking about. This work raises some doubts especially when talking about leaks with sensitive data, especially when they know the working conditions of transcriptionists, at least a good part of them: they listen to about 5,000 recordings per week per reviewer, they are not Google employees, but a third company, they have a contract for work and service or even freelance and hardly any of them reach the minimum interprofessional salary. According to workers, the projects to review what the software understands or to transcribe the user’s audio from scratch -when talking to Google Assistant- are changing: there are those of mere searches on Google, Google Maps, messages and even requests to Google Home. These language experts are employees of a technological company subcontracted by Google, with which they have to sign a confidentiality contract. Among them, there are those who still worked in this type of tasks until last week, when the scandal broke and all projects were suspended, and other former transcribers who devoted themselves to that work for years. In the absence of an official investigation, what seems to exist is a transparency problem, in the sense that the user is not aware of what is being done or what could be done from their interactions with the virtual assistant. According to consulted transcriptionists, they all agree that most of the recordings are initiated by the ‘Ok Google’ command. “In cases where the microphone stops working, we could not transcribe what is said. Sometimes you hear: ‘Google, this is not for you’. But others do not realize and continue their conversation, which is heard in the audio, although we did not write it. We also could not write down when they say personal documents, identity, banking and others. What we do have to transcribe are the mobile phone numbers and addresses, explained several former employees of the firm who prefer to remain anonymous. Google argues that the audio fragments are not associated with user accounts as part of the review process. “The system only shows you the audio, its duration and a section where you can write or review what has already been written. We do not have access to any personal data beyond what is heard,” say the language experts. However, this does not mean that the company cannot identify the data with specific users. Spokespeople from Google did not want to make clarifications about these practices and referred all questions to the official response issued by the company. But what type of searches or conversations do these transcribers get? The answer is everything that anyone can search on the Internet. There are family conversations, among friends, discussions and even intimate talks by couples. Also a lot of consultation of addresses and places or stores, as well as sexual messages and pornographic web searches. However, the privacy of users is questioned, since in the audios sensitive data is heard in an environment that Google does not control completely. In the cases of audios pertaining to interactions with Google Home, the activation of the system also marks the ‘Ok Google’. However, on many occasions the recording begins without this warning. It happens a lot that the person says something similar and the microphone is activated. Some realize and even get scared, although others continue with their personal conversation without knowing that it has been activated. In these recordings, what is mainly heard are requests for a connected house: Above all to turn on or off the light, put on music, open the garage, turn on the air conditioning and similar commands. That is what we can prove that Google is doing right now. It is likely that future leaks will expose more concerning behavior later on. After all, Google did not tell anyone they were listening. They only admitted it after the leak was made public. More news from The Real Agenda: http://www.real-agenda.com