Facebook Listened to Messenger Conversations
- August 16, 2019
- Clayton Rice, K.C.
Google. Apple. Microsoft. Amazon. And now, Facebook is the latest tech giant to confirm using outside contractors to transcribe clips of audio from users of its services. Here’s the story.
On August 13, 2019, in the lead article carried by Bloomberg News titled Facebook Paid Contractors to Transcribe Users’ Audio Chats, Sarah Frier reported that the company had “paused human review of audio” affecting users of Facebook’s Messenger app who opted to have their voice conversations transcribed. The contractors were checking whether Facebook’s artificial intelligence correctly interpreted anonymized messages. The report came after it was learned that the other four companies used human contractors to listen to audio obtained by their voice assistant products without explicit transparency.
Facebook Messenger has offered a feature to transcribe voice to text since 2015. It is turned off by default. In a post to The Verge titled Facebook also hired human contractors to listen to audio from its Messenger app Natt Garun reported that Facebook claimed “only those who opted in to the feature had their audio clips reviewed by third party contractors.” But, according to Facebook’s support page, if one person in a chat consents to transcription “any audio in the thread would have been translated, regardless of who sent it.” And nowhere does Facebook indicate humans would be reviewing the audio.
The companies all claim that the recordings are anonymized and the reason for listening to them is that the artificial intelligence used in devices like Amazon’s Echo don’t understand everything humans say. But, that is really a claim of false anonymity. Users may be recorded giving out personal identifying information even if the recording is anonymous.
In a post to Slate titled Do Tech Companies Really Need to Snoop Into Private Conversations to Improve Their A.I.? dated August 14, 2019, April Glaser asserted that the privacy invasion here is more insidious than simply having another human listen to our recorded conversations. It is one thing for Facebook to “pause” reviewing the recordings – it is quite another to continue recording in the first place. According to Meredith Whittaker, co-director of AI Now Institute, “[t]he focus on privacy needs to be about how we get in a situation where these devices are littering our homes and public spaces and we’re ultimately contributing sensitive data to corporations whose incentives and motives may be significantly different than ours”.
Littering paints the right image. The sales statistics of home assistant devices in 2018 were staggering. Amazon was estimated to have sold more than 100 million devices including Echo and Dot that feature Amazon’s digital assistant, Alexa. Apple sold 86.2 million Siri-enabled Homepods and Google’s Home sales generated $3.4 billion in revenue that included smart speakers, Nest cameras and thermostats. (See: Dieter Bohn. Amazon Says 100 Million Alexa Devices Have Been Sold – What’s Next? The Verge. January 4, 2018; Malcolm Owen. HomePod sales up in fourth quarter, Amazon and Google extending lead. AppleInsider. February 19, 2019; and, Jillian D’Onfro. Google’s small hardware business is shaping up, could book $20 billion in sales by 2021, RBC says. CNBC. December 21, 2018)
Voice assistants can be accidentally activated by ambient noise, or sounds used as “wake words”, that can trigger transmission and recording without a direct command. Accidental activation reportedly led to some of the most egregious violations of privacy as contractors reported hearing personal identifiable information, job interviews, medical consultations and couples having phone sex. (See: Alex Hern. Facebook admits contractors listened to users’ recordings without their knowledge. The Guardian. August 14, 2019)
The Information Commissioner’s Office in Britain has started an investigation into whether Facebook breached the General Data Protection Regulation (GDPR) of the European Union. Ireland’s Data Protection Commissioner (DPC), Facebook’s regulator in the EU, has also initiated an inquiry. “Further to our ongoing engagement with Google, Apple and Microsoft in relation to the processing of personal data in the context of the manual transcription of audio recordings, we are now seeking detailed information from Facebook on the processing in question and how Facebook believes that such processing of data is compliant with their GDPR obligations,” the DPC said. The DPC already has eight ongoing investigations into the social media behemoth, two into its WhatsApp subsidiary and one into Facebook-owned Instagram. (See: Irish regulator queries Facebook on transcription of users’ audio. Reuters. August 14, 2019)
In the United States, lawmakers have not all been bystanders on the privacy front. On July 25, 2019, Massachusetts congressman Seth Moulton (D) introduced a bill limiting how smart devices collect data. The Automatic Listening and Exploitation Act (Alexa Act) would authorize the U.S. Federal Trade Commission (FTC) to issue monetary penalties up to $40,000 for each infraction if a smart device records a user’s conversation without the “wake word” being triggered. Customers who use the devices would also be able to request deletion of any voice recordings, transcripts or videos collected.
“Smart speakers and doorbells are great, but consumers should have a way to fight back when tech companies collect more data than Americans have agreed to give up,” Mr Moulton said. “More broadly,” he continued, “Congress should give Americans a bigger say in the data that companies collect. It’s time for a next generation of digital privacy laws, and it can start by holding corporations to their own privacy commitments.” (See: Makena Kelly. Seth Moulton tackles Alexa data collection with new bill. The Verge. July 24, 2019)
Parliament should give Canadians a bigger say, too.