White-collar gig economy? Startup wants workers to train AI to replace them

Mercor is hiring white-collar workers who are struggling with employment to train artificial intelligence (AI) models to perform their old jobs. In the future, these models could further reduce their employment possibilities.

San Francisco-based startup Mercor has hired over 30,000 contractors in 2025 to train AI models for major companies, including OpenAI and Anthropic, the Wall Street Journal reports.

The company seeks to hire astronomers, psychologists, industrial engineers, creative writers, and other white-collar workers with a high level of expertise.

The pay ranges from $45 an hour for a social-media marketer who writes video captions to $250 an hour for a dermatologist who helps a healthcare partner develop its “decision-support tools.”

A poet enhancing AI models’ “poetic structure, literary nuance and emotional expression” may earn as much as $150 an hour.

Mercor contractors who spoke to the WSJ said they had mixed feelings about the job, as the models they train could lead to more job losses in the future.

Unemployment has been on the rise in the United States, with 2025 marking the worst year for hiring since the COVID-19 pandemic in 2020.

It remains unclear how much job loss can be attributed to AI. Recent data suggests the major reason behind layoffs remains “market and economic conditions,” although companies try to frame them as due to automation to mask their business failures.

Curious what others think about this story? Contribute your thoughts to the debate below.

The company’s spokesperson told the WSJ that humans won’t run out of meaningful work, and many of its contractors see it as their responsibility to “infuse their knowledge and expertise into the models to ensure accurate and thoughtful outcomes.”

While the technology hasn’t yet shaken up the job market, some forecasts are bleak. A recent MIT study estimated that AI can already replace 11.7% of the US workforce.

The entry-level jobs may be hit the hardest, as over 40% of business leaders are planning to replace these roles with AI.

“Gig economy of expertise”

Gad Allon, a professor at the Wharton School of the University of Pennsylvania, wrote in his Substack post that Mercor is described as building a “gig economy of expertise,” similar to Uber for transportation.

He argued that while highly skilled contractors may enjoy premium rates on Mercor, those performing more routine labeling or evaluation work can face instability and downward pressure on their pay.

Precariousness isn’t a new phenomenon in the AI data training industry. There are millions of data workers in the Global South who work on labeling, tagging, and transcribing data, as well as content moderation.

Among them, reports of grueling conditions are common, as workers of so-called “digital sweatshops” are working 20 hours a day or are exposed to large amounts of graphic and hateful content while being paid the bare minimum.

While conditions at Mercor’s differ from those of data workers in the Global South, the company was accused in 2024 of collecting personal information from interview conversations to train its AI models.

The author of a viral LinkedIn post wrote, “These companies record your interview videos and collect data from the forms you fill out, using this information to train their AI models rather than offering real job opportunities or the promised compensation.”

Unlock more exclusive Cybernews content on YouTube.