Selling Your Phone Calls, Voice, and Videos to Train AI: The Gig Apps, the Pay, and the Risks in 2026
A new gig economy is booming in 2026, and the product being sold is you — your phone calls, your voice, your face on video, even footage of you folding laundry. Companies desperate for high-quality human data to train AI models are now paying ordinary people through gig apps to hand over their most personal recordings. What started with niche data-labeling platforms has exploded into a multi-billion dollar industry where DoorDash drivers film their dishwashing routines and teenagers sell recordings of their private phone conversations.
The New Data Gold Rush
Apps like Kled AI, Silencio, Neon Mobile, Luel AI, and dozens of others are paying users small amounts for data that AI companies desperately need. As Silicon Valley’s hunger for high-quality, human-grade training data outpaces what can be scraped from the open internet, these data marketplaces have sprung up to fill the gap.
- Neon Mobile — Pays users $0.30/minute for recorded phone calls (Neon-to-Neon) or $0.15/minute for one-sided recordings, up to $30/day. An 18-year-old in Chicago made hundreds selling his private conversations with friends and family. The app shot to #2 on the Apple App Store in September 2025.
- DoorDash Tasks — Launched March 2026, this standalone app pays DoorDash’s 8 million U.S. couriers to submit videos training AI and robotic systems. Tasks include recording unscripted Spanish conversations, filming household chores like loading dishwashers and folding clothes, photographing restaurant menus, and capturing hotel entrance footage.
- Uber AI Solutions — Launched October 2025 as a pilot letting U.S. drivers complete AI data-labeling tasks — uploading photos, recording audio clips, and transcribing documents — through the Uber Driver app. Already tested with over 1 million drivers across 12 cities in India.
- Kled AI — Pays users for uploading photos and videos of everyday life. One contributor in South Africa earned $50 in two weeks by recording walks around his neighborhood.
- Silencio — Crowdsources ambient audio data by accessing users’ phone microphones. A student in India earns over $100/month recording city sounds like restaurant noise and traffic.
- Luel AI — A Y Combinator-backed (W26) data marketplace acquiring multilingual dialogue samples at ~$0.15/minute. Delivers licensed, audit-ready data with GDPR compliance documentation.
- Mercor — Pays professionals $45–$250/hour (average $95/hr) to share expert knowledge. Doctors earn $130–$170/hr, lawyers $110–$130/hr. The company pays $1.5 million per day to ~30,000 contractors and is valued at $10 billion.
DoorDash released a new app called Tasks. It allows the 8 million Dasher couriers in America to create video and audio content for AI robotics training data.
— Bearly AI (@bearlyai) March 19, 2026
The Full Lineup: What Each Platform Pays
| Platform | What They Collect | Pay Rate |
|---|---|---|
| Neon Mobile | Phone call recordings | $0.15–$0.30/min ($30/day cap) |
| DoorDash Tasks | Video of household tasks, conversations | Per-task (complexity-adjusted) |
| Uber AI Solutions | Photos, audio clips, transcriptions | Per-task |
| Mercor | Expert knowledge (doctors, lawyers) | $45–$250/hour |
| ElevenLabs Voice Library | Voice clones for licensing | $0.02/min (usage-based) |
| Kits AI (Kits Earn) | Vocal recordings for AI voice models | Usage-based royalties |
| Luel AI | Multilingual dialogue samples | ~$0.15/minute |
| Babel Audio | Topic-specific conversations | ~$17/hour |
| Appen / Scale AI | Speech, audio, text, image labeling | Varies (often very low) |
| DataAnnotation.tech | AI output ranking, LLM evaluation | Starting ~$20/hour |
Who’s Buying Your Data — And How Much They’re Spending
The buyers behind these data marketplaces are the biggest names in tech, and they’re spending staggering amounts:
- Meta invested $14.8 billion for a 49% stake in Scale AI in June 2025 — roughly 25% of Meta’s entire $65 billion projected AI budget for the year. Scale AI is the largest data-labeling operation, with over 1 million contributors in 170 countries.
- Google spent approximately $150 million on Scale AI services in 2024 alone, though it’s reportedly pulling back after Meta’s massive investment.
- OpenAI is working with Juilliard music students to train composition models and former investment bankers for Wall Street-specific tasks. The company also acquired io Products (Jony Ive’s design firm) for ~$6.5 billion to build voice-first AI hardware.
- Robotics companies (unnamed) are purchasing DoorDash’s task video footage specifically to train humanoid robots in manipulation tasks like dishwashing and clothes folding.
The global AI training data market is valued at $3.59 billion in 2025 and projected to reach $23.18 billion by 2034, growing at nearly 23% annually, according to Fortune Business Insights.
Why AI Companies Are Desperate for Your Data
AI language models like ChatGPT and Gemini are facing a genuine data drought. Researchers at the World Economic Forum warned that the total stock of human-generated internet data could be depleted as a usable training source by 2026. The most-used training sources — Wikipedia, Reddit, news archives — are now restricting access or demanding licensing fees.
And feeding AI its own synthetic data? Researchers describe it as “the computer-science version of inbreeding” — it leads to “model collapse” that produces increasingly nonsensical outputs. Real human data — authentic conversations, real-world video, genuine audio — remains the gold standard for training AI that can actually function in the real world.
This is why companies are willing to pay gig workers to record themselves doing mundane tasks. A 30-second video of someone loading a dishwasher is worth more to a robotics company than a thousand stock photos.
The Privacy Nightmare Nobody Reads the Fine Print For
The trade-offs of selling your personal data to AI companies go far beyond a bad deal — they’re a privacy catastrophe waiting to happen.
Neon’s Security Breach: In September 2025, just days after reaching #2 on the App Store, TechCrunch discovered that Neon’s servers had a flaw allowing any authenticated user to access everyone else’s phone numbers, call recordings, and transcripts. The app went offline for a month before returning with stricter app-to-app requirements.
Irrevocable Licensing: Neon’s Terms of Service grant the company an “exclusive, irrevocable, and transferable license to sell, modify, and distribute recordings.” Once you upload your voice, it’s no longer yours — legally or practically.
Two-Party Consent Violations: About a dozen U.S. states — including California, Florida, and Pennsylvania — require all-party consent to record phone calls. Apps like Neon may be violating these wiretapping laws every time a user records a call with someone who hasn’t installed the app.
Deepfake Risk: When you sell a recording of your voice or face, you’re giving AI companies the raw material to create synthetic versions of you. A New York actor sold his likeness for $1,000 with usage restrictions, only to discover his friends forwarding AI-generated videos featuring his face and voice that had amassed millions of views.
The Lawsuits Are Piling Up
The legal reckoning for AI data collection is already underway, and the lawsuits are multiplying fast:
- Scale AI / Outlier AI — Hit with multiple lawsuits in December 2024 and January 2025 for underpaid wages. Workers report being paid for only half their hours, with effective pay as low as $0.01 per task. The Oxford Internet Institute scored Remotasks (Scale AI’s platform) 1 out of 10 on fair work criteria.
- Otter.ai — Federal class-action filed August 2025 alleging the transcription app secretly records private work conversations without participant consent.
- Fireflies.AI — Sued in December 2025 for allegedly recording, analyzing, and storing voiceprints of meeting participants without consent.
- Cresta Intelligence — Alleged monitoring of United Airlines calls without consent, violating California’s Invasion of Privacy Act.
- BIPA Settlements — Under Illinois’ Biometric Information Privacy Act, Clearview AI paid $51.75 million, Motorola FaceSearch paid $47.5 million, and over 107 new BIPA class actions were filed in 2025 alone.
Regulators are also stepping in: the FCC mandated written consent for AI calls starting January 2025, and the EU AI Act’s full applicability begins August 2026, requiring disclosure of training data sources.
The Workers Training Their Own Replacements
Here’s the darkest irony of this entire gig economy: the people selling their data are actively training the AI systems that will replace them. DoorDash drivers filming themselves doing household chores are teaching robots to do those same chores. Voice actors selling recordings are building the datasets that make AI voice cloning good enough to eliminate voice acting jobs.
As one TechFlow investigation put it: “These trainers are fueling an industry that may ultimately render their own skills obsolete — and exposing themselves to future risks of deepfakes, identity theft, and digital exploitation.”
The pay disparity is also telling. While Mercor pays doctors and lawyers $130–$250/hour for their expertise, platforms like Appen and Scale AI’s Remotasks pay gig workers in developing countries as little as a fraction of a cent per task. Privacy International has documented how the data labeling process is “deliberately obfuscated” by companies that benefit from keeping workers invisible.
The Bottom Line
The logic is grimly rational: tech companies already harvest your data without paying you, so why not get paid? But there’s a difference between a company tracking your browsing habits and literally selling recordings of your private phone calls to the highest bidder.
This isn’t a side hustle — it’s selling your identity one data point at a time. The $14 you earn recording your feet while walking might be the cheapest your data ever sells for. And unlike driving for Uber or delivering for DoorDash, you can’t take back a voice recording once it’s been used to train an AI model that can clone you.
The AI training data market will be worth $23 billion by 2034. The question is whether the people generating that data will see anything close to a fair share — or whether they’ll be remembered as the generation that sold their voices for pennies while tech companies built trillion-dollar empires on top of them.