AI data training scandal: App that paid users to record calls for AI training goes offline after major security lapse

Alex Kiam, Neon’s CEO and a Stanford MBA graduate, confirmed that the service will remain offline for at least one to two weeks while undergoing a full security audit. He said additional safeguards, including row-level protections, will be added before the app returns. Kiam admitted the pace of Neon’s rise had taken the small four-member team by surprise, describing the surge from hundreds to tens of thousands of users in days as “faster than expected.”

Launched earlier this month, Neon allowed users to make and receive calls via its in-app dialer. The app paid 30 cents per minute when both parties were Neon users, and 15 cents when only one caller used the service. Users could earn up to $30 daily, with payouts processed within three business days. While the company said call recordings were anonymized before being sold, Kiam acknowledged that no agreements with AI firms had yet been finalized.

The bigger picture

The development has intensified scrutiny of how far companies are prepared to go in sourcing training data for artificial intelligence models. U.S. laws differ on whether one or both parties must consent to recording, leaving ambiguity around compliance for a service like Neon. While the company insisted it had designed systems to comply with these laws, revelations of exposed call data have fueled debate over whether consumer-facing platforms can balance profit incentives with user privacy.

As major firms race to expand AI training resources, the Neon episode underscores the risks of commercializing personal conversations, highlighting how demand for large datasets can collide with privacy and security obligations.