Advertisement

SKIP ADVERTISEMENT

No, Facebook Is Not Secretly Listening to You

(Except when it is.)

Credit...Getty Images

Ms. Jeong is a member of the editorial board.

This article is part of a limited-run newsletter. You can sign up here. Charlie Warzel is off this week, so he’s turning the newsletter over to his colleague Sarah Jeong.

Ever have an uncanny moment on Facebook or Instagram when you saw an advertisement for something you were just talking about in real life? You never searched for it on your phone or typed the words on your computer, so how could advertisers possibly know? Was your phone bugging you?

No, your phone is not secretly listening to you.

Last year, Kashmir Hill (now doing investigative work here at The New York Times) wrote about an academic study at Northeastern University that looked for “unexpected activity” transmitted from smartphones — in other words, an audio file secretly broadcast to Facebook or its advertisers. She wrote:

They found no evidence of an app unexpectedly activating the microphone or sending audio out when not prompted to do so. Like good scientists, they refuse to say that their study definitively proves that your phone isn’t secretly listening to you, but they didn’t find a single instance of it happening.

The study included apps belonging to Facebook and over 8,000 apps that send information to Facebook. Those apps came up clean. (The study did detect some untoward behavior from other apps — some were recording screenshots and video of what people were doing on their phones and transmitting them to third-party analytics companies. But none of them was listening in on conversations).

This is my personal bugbear, the thing that sets me off on a probably tedious tirade to my friends and neighbors at backyard barbecues and cocktail parties. The simple fact is that your devices gather so much data about you, your whereabouts, your contacts, your browsing activities and the activities of your contacts that advertisers can predict what you’re saying without even actually listening in.

Unfortunately, my insistence that Facebook is not listening to you is, predictably, undermined by Facebook, which sometimes is secretly listening to you.

Last week, Bloomberg reported that hundreds of outside contractors have had access to audio chats in Facebook Messenger — because they were paid to transcribe them. Of course, these conversations were happening over Facebook Messenger, rather than being passively recorded in the privacy of one’s home. And they happened only when the user opted in for message transcription.

But the opt-in message didn’t mention any human beings. “Turn on Voice to Text in this chat?” the pop-up asks in bold text. Below that, it reads: “Display text of voice clips you send and receive. You can control whether text is visible to you for each chat. Learn more in the Me tab.”

If this latest admission from Facebook sounds familiar, it’s because this is the fifth big tech company recently revealed to be using human ears when you might not expect it. In April, Bloomberg reported that Amazon Alexa uses human beings to review audio files to check whether the technology is correctly interpreting voice commands. In July, a human contractor for Google Assistant leaked thousands of audio clips to a Belgian public broadcaster. The MIT Technology Review reported: “The clips include fragments of deeply personal conversations, including people’s addresses, information on someone’s love life, and what sounded like a woman in distress. Many of the recordings were captured accidentally, because the speaker had incorrectly identified the ‘wake word.’”

The same month, a whistle-blower told The Guardian that Apple uses contractors to listen to Siri audio clips. Shortly after that, Vice reported that Microsoft contractors listen to Skype conversations that use Translator, a seemingly miraculous feature that provides real-time audio translation during calls.

In all five of those cases, the contractors are being used for quality control. The artificial intelligence that drives voice-to-text or automatic translation can improve over time only when a human being flags when it succeeds or fails. But this drudgery is sometimes a shockingly prurient invasion of privacy. “Apple contractors regularly hear confidential medical information, drug deals, and recordings of couples having sex,” The Guardian reported in its story about Siri.

Because voice assistants like Siri, Alexa and Google Assistant are triggered mistakenly all the time, people are being recorded without their knowledge. It doesn’t quite amount to wiretapping, because they’ve consented to having these devices in their lives. But I highly doubt that users reasonably expect recordings of their sexual activities to be played back to other human beings who are being paid to listen.

This wouldn’t be so bad if the companies were explicit about what was going to happen with these audio files. And the blow would also have been diminished if they hadn’t oversold the power of artificial intelligence to people who weren’t in a position to know better.

Voice recognition is still in its infancy. As it turns out, the human brain makes a lot of leaps and inferences with the signals it receives through the ears — assumptions and fill-ins that we’re still trying to figure out how to do with machines. Artificial intelligence needs to feed on data, and in this case, the data takes the form of these personal audio files. But it’s not enough to just dump the audio into a machine. Human contractors must be paid to recognize what machines cannot.

Being upfront about the humans who operate behind a curtain of artificial intelligence would mean looking less ingenious, less innovative, less omniscient. But users deserve to be in the know and able to make informed decisions about what devices to allow in their homes and on their persons.

In 2001, four Times reporters sat down with four “veterans” of Silicon Valley, including Eric Schmidt of Google. The dot-com bubble had just burst, so much of the round table was preoccupied with economic predictions. Some of the conversation reads like a time capsule. (One reporter asks whether they are concerned about “studies that suggest that people who don’t use the internet see no use for it.”)

Other parts are still relevant. The tech executives were grilled on whether “free” services are still viable as a business model and whether advertising is still the model for the future. Mitchell Kertzman, then chief executive of Liberate Technologies, said the current moment was seeing “an overreaction” against advertising. Mr. Schmidt — as prescient as ever — observed: “There is an enormous demand for all-you-can-eat subscription services for proprietary intellectual property. People are tired of having 500 accounts which charge them $5 a month for this source and that source.” (Netflix was founded in 1997 and introduced its streaming service in 2007.)

Mr. Schmidt also predicted that in the coming years:

… there’s clearly going to be pressure around privacy legislation of some kind. A lot of people are worried that the privacy legislation will end up being written by a large number of lobbyists from all the different sides, and it will not end up looking like whatever outcome people really think it should.

This was President George W. Bush’s first year in office, and in an alternate universe, we may very well have seen pro-privacy regulation in America enacted then. The round table took place in July 2001. After Sept. 11, the government passed the Patriot Act, an enormous erosion of American privacy that we are still contending with.

Last week, Charlie gave some great tips to stay safe on public Wi-Fi. (His explanation of virtual private network services can be found here.) Shortly after the newsletter landed in your inbox, Wirecutter, a New York Times-owned site, put out its 2019 recommendations for the best V.P.N.s. It’s well worth a read just to start thinking about what kind of standards you want to set for your V.P.N., and really, any consumer technology.

“All of your internet activity will flow through the servers of the company whose V.P.N. you use, so you’ll need to trust it more than you trust the network you’re hoping to secure, whether that’s airport Wi-Fi, a hotel internet connection, your corporate IT network, or your home I.S.P.,” says Yael Grauer, the technology journalist who wrote the recommendation. She then rattles off a series of stories where V.P.N.s made promises they did not live up to. (For example, earlier this year, “more than half of the top 20 free V.P.N.s in the App Store and the Google Play store were owned by or based in China, a country where V.P.N. services are banned.”)

Ms. Grauer assessed trustworthiness based on news reports, on information about who owns and operates the service, on the opinion of security experts and on court cases where specific V.P.N. services have proved to authorities that they don’t log customer data. She also looked for security audits by third parties.

All in all, it’s a great primer on how to think about the tech you put your trust in.

[If you’re online — and, well, you are — chances are someone is using your information. We’ll tell you what you can do about it. Sign up for our limited-run newsletter.]

The administration has asked Congress to reauthorize a law that authorized the National Security Agency to sift through records of Americans’ calls and text messages. If Congress doesn’t act, the law will expire in December.

From the annals of unintended consequences: “Researchers Show How Europe’s Data Protection Laws Can Dox People.”

Three Years of Misery Inside Google”: This account dives into details about employee activism inside the company, including activism against what employees believed to be unethical uses of artificial intelligence. It’s well worth a read. But long story short? Many of the employees who led the charge believe they have been retaliated against and have left the company.

Like other media companies, The Times collects data on its visitors when they read stories like this one. For more detail please see our privacy policy and our publisher's description of The Times's practices and continued steps to increase transparency and protections.

Follow @privacyproject on Twitter and The New York Times Opinion Section on Facebook and Instagram.

glossary replacer

Advertisement

SKIP ADVERTISEMENT