The AI Gig Economy Is Here—And It Pays in Crypto

Sahara AI’s new bounty platform rewards users for labeling training data, with over 50K in token payouts

By Jason Nelson Edited by James Rubin Jul 22, 2025 Looking for a way to earn crypto without writing code or trading tokens? Try labeling data for AI. Los Angeles-based Sahara AI debuted its Data Services Platform (DSP) today, a new system that pays users in cryptocurrency for completing small but essential data-labeling tasks—tagging images, transcribing audio, or evaluating AI-generated text. The platform aims to turn the grueling, often underpaid work of training AI systems into a decentralized gig economy. “We've been running some closed beta version of it in the past few months, and now we are moving into this open version, where everyone can come to our platform to check on the data tasks that are published by our ecosystem partners,” Sahara AI co-founder and CEO, Sean Ren, told Decrypt. The platform is modeled on bug bounties where companies, research labs, and crypto projects post data tasks that contributors complete in exchange for rewards. Instead of finding software bugs, though, DSP users get paid to annotate data. According to Sahara, more than 50,000 in crypto rewards are available at launch. Ren, who also teaches computer science at the University of Southern California, said the bug bounty model helps ensure better data quality while discouraging bad actors through methods like automated checks, peer review, contributor reputation scores, and requiring users to stake tokens as a defense against attacks. Sahara follows other crypto-native bounty models, such as Immunefi, which pays out for security vulnerability disclosures, and Gitcoin, which supports open-source software development through grants, bounties, and funding. According to Sahara, DSP tasks fall into three categories:

• Enterprise Tasks, which pay in tokens for helping major clients structure or label data.
• Dual-Reward Tasks, which pay in and ecosystem partner tokens.
• Community Tasks, which don’t offer immediate payouts but give contributors ownership stakes in resulting datasets and recurring income when those datasets are used or sold.

Sahara also supports rewards in stablecoins, including USD1, USDC, and USDT as other reward options. While bug bounties aren’t new, Ren said the model focuses more on providing quality data, not just fixing holes. “Data labeling is fundamental for training accurate AI models,” he said. “Labeled data acts as a guide to teach models to recognize patterns and make predictions.” To maintain data quality, Sahara employs several security and review mechanisms. “We monitor the performance of data labelers using honeypot questions,” said Ren. Those are designed to detect users attempting Sybil attacks, where one person has multiple fake identities to game the system. “Cheaters can be quickly identified with our quality assurance mechanisms and then be temporarily banned or long-term banned,” he said. Likewise, the company is using AI to catch cheaters—ironically, using AI to find users who are themselves using AI to create content. –- Source: Originally published at Decrypt on Jul 22, 2025.