The head of Instagram swears the app isn't listening to you. But here's what it is doing. (www.businessinsider.com)

Instagram head Adam Mosseri posted a short “myth-busting” video denying the app listens to users’ real-world conversations to serve ads, calling microphone surveillance a privacy- and battery-draining implausibility. He explained targeted ads instead come from signals such as what you tap or search for, advertiser-shared data (e.g., website visitors tracked by pixels), cross-platform activity, third‑party cookies, friends’ interests and lookalike behavior, and even prior exposure to the same ad. Mosseri also acknowledged coincidence can make ad timing feel uncanny, and reiterated that visible mic indicators would alert users if audio were being recorded. The denial matters less for advertisers than Meta’s simultaneous announcement that interactions with its AI chatbots will soon be used to personalize content and ads—users will be notified next week and the change takes effect December 16, 2025. For the AI/ML community this highlights two trends: more signals (including conversational AI interactions) will feed recommendation and ad models, and platform-collected AI-chat data may be repurposed for personalization and training. That raises technical and ethical implications around data provenance, consent, notification, retention policies, and the role of human annotators (reports say contractors have reviewed AI chats), which affect model auditing, privacy engineering, and regulatory compliance.