The AI data industry runs on a labor model it inherited from gig platforms: thousands of workers, engaged as independent contractors, doing scheduled, supervised, rubric-governed work that looks a great deal like employment. Worker-classification lawsuits have now arrived in this industry — class actions included — and the legal theory is not exotic. It's the same one that reshaped rideshare.
Why labs should care about their vendor's labor model
Misclassification exposure doesn't stay neatly inside the vendor. If the vendor's model collapses, your data pipeline collapses with it. If a regulator asks who is legally responsible for the person who graded your safety evals, 'a 1099 contractor of a subcontractor' is not an answer anyone wants to give under oath. And joint-employment theories exist precisely to pull the client into the frame.
- Scheduled shifts, mandatory training, and detailed work rules are classic employment indicators.
- Cross-border contracting multiplies the problem: every country draws the employee line differently.
- Indemnities help until the counterparty holding them can't pay.
The boring, correct alternative
Employ the workforce. Properly, in-country, with payroll, tax, and benefits handled by an entity built for it. That's the infrastructure we've operated for over a decade across 180 countries — it's not a pivot for us, it's the day job. Workers get real employment; clients get a workforce whose legal responsibility sits with us, in writing.
Compliance isn't a feature we added to a labeling company. Labeling is a workload we added to a compliance company.
When the next classification ruling lands, the question every lab will ask its vendors is 'who employs your people?' We like our answer.
We were the employer