Gig workers selling personal data for AI training are sensing opportunity but in reality, are they making themselves obsolete?


All your AI-powered chatbots like ChatGPT and Claude depend on data for training. But as per research, they will run out of content to train on as early as mid-2026.

That’s where Priyanka (23) and Laveena (19) come in. Based out of Ranchi and Kanpur respectively, they document their everyday life, sharing photographs, audio clips and video clips to apps like Kled AI or Neon Mobile, that pay contributors for uploading their data to train artificial intelligence models.

And they’re not alone, Nikumbo Dondangi (27) in Kenya does the same thing. He records audio clips of himself as well as ambient sounds around him and sells them to an app called Silencio, that crowdsources audio data for the same training purpose.

Of course, they get paid for this. Priyanka earns around $10 for a video while Nikumbo makes over $100 a month. But these websites are shifting how things work in the AI gig economy.

From Labour to Likeness

As far as these gig workers are concerned, there is nothing wrong in what they do. Without mucheffort, they’re making a quick buck; in fact, a lot more than they can hope to earn otherwise.

The problem however lies in the fact that they’re selling a lot more than just audio or video clips. They’re selling a license on their data which can be used “worldwide, exclusive, irrevocable, transferable and royalty-free.”

A 20-minute video is enough to train a customer chatbot for many years. The original video provider would never know about it or earn more than he did for that single video. Similarly, sold photographs can be entered into facial recognition software without their explicit consent.

Videos and audio clips can be used to generate deepfakes or advertisements in another country or continent. The original content providers would never even find out or benefit from these AI-generated clips.

The Illusion of Consent

As far as the gig workers are concerned, they are getting paid for their content. But they don’t realize that this is shaped by economic pressure.

Most of these AI trainers are from developing countries where any kind of earning is a win, especially in today’s economy. And if that earning is in US dollars, it is a major win.

The problem here is that consent in this context is very vague. The users agree to terms and conditions before selling their content but it is not clear where or how their data will be used. Once the data enters an AI system, there is no controlling how it will be used.

And there are no current laws that exist that will help you take legal action against illegally using your data that you explicitly sold.

A Race to the Bottom

Like almost all gig economies, this one is following a similar pattern. The income is low. It is also irregular. And there is no protection of any sort.

With so many people across the world willing to sell their data for any price, it has also driven the prices to rock bottom. Personal data is increasingly being sold for diminishing returns.

And the biggest paradox of them all is that, the very data that they are selling will make them expendable in the future. By using this data from human to train AI, the humans will become obsolete.

The Last Word

The big question here is who really owns artificial intelligence when it is built on human lives? The companies call it innovation; the workers think it is opportunity. But in reality, the platforms are making long-term gains while the works have a one-time payment. And it isn’t just about the effort – it is identity, memory and behaviour.

The AI revolution is a technological leap for sure. But once the transfer of human experience into machine systems is complete, there will be no going back.

In case you missed:

Adarsh hates personal bios, Chelsea football club and Oxford commas. When he's not writing, he's busy playing FIFA on his PlayStation.

Leave A Reply

Share.
© Copyright Sify Technologies Ltd, 1998-2022. All rights reserved