Investigation details how Amazon, Apple, Google, and Facebook handle voice assistant transcriptions, relying heavily on contractors and downplaying user privacy
This investigation is the capstone of a year-long drip of revelations rather than a new disclosure. It started with leaked Google Assistant clips obtained by a Belgian broadcaster in July, followed by Google's admission that one of its language reviewers had leaked confidential Dutch audio, then Facebook's confirmation that it had paid hundreds of contractors to transcribe user voice chats. By late August the pattern had already drawn regulators: Germany's data protection authority was probing Facebook over the practice alongside its existing investigation into Google's speech assistant.
First-order effects
The story widens the blast radius from two companies to four: Amazon and Apple are now publicly tied to the same contractor-transcription model that already put Facebook and Google under German regulatory investigation.
Each company's privacy marketing — the core selling point for always-listening devices in consumers' homes — is directly contradicted by the reporting, forcing immediate clarification of what consent disclosures actually cover.
Second-order effects
Regulators with open probes into Facebook and Google now have a template covering Amazon and Apple, making a coordinated European response across all four assistants more likely than piecemeal national actions.
Contractor staffing for AI annotation becomes a reputational liability for the whole supply chain: vendors doing this work for one assistant face scrutiny for all their clients, and companies may rush to reclassify or curtail manual review programs as Facebook already did.
Third-order effects
If GDPR-class enforcement spreads from the two exposed companies to the entire voice-assistant market, training-data pipelines built on human review of real user audio become structurally untenable, pushing the industry toward techniques like the voice-anonymization approaches Slate reported on, or on-device processing that never sends raw audio to the cloud at all.
The trend: Voice-assistant privacy is moving from per-company scandal to an industry-wide regulatory reckoning over how human-reviewed audio trains AI.
NEW COVER: Amazon, Apple, Google, Microsoft, and Facebook are listening to your most intimate moments. Here's how the world's biggest companies got millions of people to let temps analyze some very sensitive recordings https://bloom.bg/2RHuqd2 pic.twitter.com/1EY9o6U2Fr
NEW @BW: Apple, Amazon & other co's created vast networks of humans listening to our Siri + Alexa audio. Tech execs say this isn't problematic or even intrusive. Inside the Silicon Valley's Eavesdropping Economy: https://www.bloomberg.com/... pic.twitter.com/nDfkY6bkDm
""It's not like the machine has an intention to record people making certain kinds of sounds. It's like a statistical fluke." #MutableMadness2019 Words Mean Things https://www.bloomberg.com/...
The homogeneity and culture of businesses can't be seperated from their policies and the refusal to look at it is biting us in the behind https://twitter.com/...
The recordings “were often intense, awkward, or intensely awkward. Lonely sounding people confessing intimate secrets and fears: a boy expressing a desire to rape.” As Alexa's popularity grew, “so did the private information revealed in the recordings.” https://www.bloomberg.com/…
Ruthy Hope Slatis couldn't believe what she was hearing. For $12 an hour, she and others listened to snippets of conversations and jotted down every word. Amazon would only say the work was critical to a top-secret speech-recognition product https://www.bloomberg.com/...
“It never felt right,” says a voice transcriber for an Alexa rival who, like most of the contractors, signed a nondisclosure agreement and spoke on condition of anonymity for fear of reprisals. “What are they really selling to customers?” https://www.bloomberg.com/...
The main thing we all realized when reporting out this story: - The contractors listening to voices were all horrified, and saw an obvious violation - the companies were all surprised by the discomfort. Listening to audio to improve their AI was a natural next step for the tech. …