They say talk is cheap. But Neon Mobile briefly put a price on it. The app surged to the second spot of Apple’s U.S. App Store by paying users to record phone calls. Those conversations are then sold to artificial intelligence companies for training purposes.
However, a security flaw in the company’s technology exposed private information, leading to the service going dark.
Download the SAN app today to stay up-to-date with Unbiased. Straight Facts™.
Point phone camera here
How Neon Mobile works
According to its website, anyone can download the app for free, and Neon pays users every time they talk on the phone.
“Phone companies profit off your data,” their site reads. “Now, you can too.”
The app records a user’s audio and then sells it to AI firms, but claims they do not record the audio of non-Neon participants. CEO Alex Kiam told Business Insider that personal information is not seen by these AI companies.
“If anything, it’s too aggressive,” he said. “There’s stuff that it bleeps out that isn’t personally identifying and is actually kind of valuable to the AI companies. But we would rather… err on the side of just being fully anonymized.”
When speaking with a non-Neon user, the pay rate was 15 cents per minute. That jumps to 30 cents per minute if both parties use Neon, with earnings capped at $30 a day. Users can also receive a $30 referral bonus for referring new people to the app.
The company’s terms of service say recordings were used “for the purpose of developing, training, testing, and improving machine learning models, artificial intelligence tools and systems, and related technologies.”
Security flaw ends the streak
The hype ended quickly. TechCrunch reported on Thursday that a flaw allowed outsiders to access sensitive data, including phone numbers, call transcripts and recordings.
Neon immediately went offline for a security audit. Kiam told Business Insider that his team would not relaunch until the problems were fixed and a security audit was conducted.
The incident highlights the intense pressure AI companies are facing to acquire new data. Large language models are already trained on massive text sets, but those datasets have limits. Human voices may be the next frontier.
A recent example is an issue Lionsgate has been running into. The Wrap reported that the studio realized its own library wasn’t enough to create a significant AI-powered movie. The production house’s shortage underscores the industry’s hunger for additional data sources.
