Handshake contractors claim $64,000 in unpaid AI-training wages

Nearly two dozen current and former contractors say the AI-training startup Handshake underpaid them by hundreds or thousands of dollars each this year, with disputes over the past six months totaling $64,000, according to the contractors, who spoke to Business Insider. The company, which says more than 188,000 contractors have earned over $500 million on its platform, has introduced pay-transparency changes after the report.

At the center of the investigation is Joel Thomas, a Virginia college student who says he is owed about $11,000 in back pay for hundreds of hours of AI-training work. Handshake invited him in August to apply for projects paying $50 an hour or more — rates he figured would cover a full year of college expenses. Instead, he says he received $8,000 for work his dashboard valued at $19,000.

What went wrong with the pay?

Business Insider reviewed bank statements, pay stubs and hundreds of email and Slack messages between contractors and the company. Several failure points emerged:

• Contractors' dashboard figures were estimates based on manually logged timer hours, while the client ran a second, invisible timer that Handshake checked before paying.
• Handshake didn't tell contractors what counted as billable time or how clients calculated it, former employees said.
• Tasks that failed review went unpaid, and project managers — sometimes managing up to 1,000 contractors on marquee projects — didn't have time to answer clarifying questions.
• Project managers used AI coding tools like Claude and Lovable to generate instructions that were sometimes contradictory, former employees said.

"Final pay reflects the work our systems verify," a Handshake spokesperson said. "Where logged time doesn't match the work, we review it, and where we find inflated time we don't pay for it."

How did the company respond?

Business Insider first reported in March that dozens of contractors complained of denied payments, and one contractor won a $6,475 judgment in small-claims court. In October, after BI sent detailed questions, Handshake introduced a payments dashboard, more explanations for adjustments, an easier dispute process and a live timer for third-party platforms.

"Our Fellows are the heart of Handshake AI," the company wrote in an October blog post. "You should always know what you earned, how it was calculated and when it will arrive."

The fixes sit awkwardly next to CEO Garrett Lord's marketing. "Most AI gigs = shifting rates, zero transparency," Lord wrote on LinkedIn last year. "Handshake AI = competitive pay, clear expectations, and real dialogue."

The student's case

Thomas, then a sophomore at Liberty University, signed up in January after the school directed him to Handshake's career platform. He took low-paid spring projects at about $17 an hour, then jumped on August offers paying $50 per hour and $65 per task. He worked from mid-morning to midnight or 1 a.m., and his dashboard topped $19,000.

Payments came in short. Handshake told him in an email that for August 17 to 23, "our project leads did not find any tasking that took place during this time frame" — even though a pay statement for that period showed 428 completed tasks on Project Novel with no pay. The company said the phrase is shorthand for "no verified work," and that it doesn't pay for unapproved tasks or time exceeding project caps.

Handshake then deactivated his account, citing irregular activity. It declined to comment on his case individually, saying only: "We take every dispute seriously."

Why it matters for workforce planners

Handshake pivoted into AI training in January 2025, and Lord said in August the unit has grown past $1 billion in gross annualized revenue — mostly from five clients: Meta, Google, OpenAI, Nvidia and Anthropic. For HR and talent teams, the case is a caution flag on gig-workforce pay practices in the booming AI-data industry: dual invisible timers, opaque review standards and thin support staffing are now documented industry-wide.

Thomas still sees AI as a career tool and remains open to training models. But he's given up on collecting from Handshake.