A new app that promises to pay people for their mobile phone call records, which is then used to train AI models, is getting so popular it’s entered Apple’s list of top-ranked free apps.

As of Thursday, Neon was the fourth most popular iOS free app, ahead of Google, Temu and TikTok. It had earlier been as high as the No. 2 spot.

Neon is available on iOS and on Android and the idea is that the company records the outgoing phone calls of users and pays them up to $30 a day for regular calls or 30 cents a minute if the call is to another Neon user. Calls to non-Neon users pay 15 cents a minute. The app also offers $30 for referrals. 

«You can cash out as soon as you earn your first ten cents,» a Neon app FAQ says, «Once redeemed, payouts are typically processed within 3 business days, though timing may occasionally be shorter or longer.»

The company promises it only draws from the recording of one side of the phone conversation, the caller’s, which appears to be a way of skirting state laws that prohibit recording phone calls without permission. While many states only require one person on a call to be aware that a call is being recorded, others including California, Florida and Maryland, have laws that require all parties on a phone call to consent to recording. It’s unclear how Neon is able to function with calls to those states. For Neon-to-Neon calls, two-party consent would presumably be implied.

The app does not record regular phone app calls, only those made within the Neon app or received from another person using Neon.

An email to Neon Mobile, the company behind the app, was not immediately returned.

While the iOS version has shot up in popularity, the Android version appears to be having some problems, at least according to some of the most recent reviews on the Google Play store. The Android app only has a 2.4 star rating and some user comments report network errors when people try to cash out on the Neon app.

According to the company’s FAQ, the call data is anonymized and used to train AI voice assistants. «This helps train their systems to understand diverse, real-world speech,» it says. 

As pointed out by TechCrunch, one of the first sites to write about the app, sharing voice data can be a security risk, even if a company promises to remove identifying information from the data.