Facebook admits it secretly listened to users' Messenger calls
Facebook secretly employed third-party contractors to listen to and transcribe private audio clips from Messenger users to train its AI, violating user privacy expectations and raising serious concerns about the nature and transparency of "automated" systems. What happened A Bloomberg investigation revealed that Facebook had been paying hundreds of outside contractors to listen to and transcribe snippets of voice conversations from its Messenger app. While Facebook claimed the practice only affected users who had opted into the "voice-to-text" transcription feature, it failed to disclose that human beings, not just algorithms, would be reviewing these private recordings. Contractors reported hearing "vulgar" and highly personal content, often without knowing where the audio came from or how Facebook obtained it. What happened The incident was driven by the technical need to "ground truth" or "grade" artificial intelligence; to improve the accuracy of automated speech recognition, human reviewers are often used to check the machine's work. However, this happened because of a significant transparency gap: Facebook’s data-use policy broadly mentioned "vendors and service providers" but did not explicitly state that humans would listen to private audio. This lack of accountability allowed the company to maintain the illusion of a fully automated process while compromising user confidentiality. What it means For users, the incident highlights that "private" communications on major platforms are rarely truly private if they are being used for AI training. For society, it exposes the "ghost work" behind AI, where low-paid human labour is used to prop up claims of technological sophistication. For policymakers, it underlines the need for stricter "human-in-the-loop" disclosure requirements and contributed to a broader wave of regulatory scrutiny regarding how Big Tech handles biometric and audio data. System 🤖 Facebook Messenger Voice-to-Text Developer: Facebook Country: USA Sector: Multiple Purpose: Transcribe voice calls; Train AI models Technology: Machine learn ing; NLP/text analysis; Voice-to-Text Issue: Accountability; Consent; Privacy/surveillance; Tran sparency Investigations, assessments, audits 👁️ Bloomberg. Facebook Paid Contractors to Transcribe Users’ Audio Chats
- Date it happened
- 2019-08-01
- Organisation involved
- Product, system or model
- Facebook Messenger Voice-to-Text
This incident was imported from AIAAIC and is used under CC BY-SA 4.0. Our additions to it — the structured fields, the translation, the checks against other reports — are published under the same licence.
This is a record of what was reported, not a finding that anyone broke the law. If it names your organisation and you believe it is wrong, the corrections process is free and open to everyone.