Mercor

Mercor confirmed that compromised LiteLLM versions affected it in March 2026, exposing a limited subset of expert information according to its June investigation update. Handling sensitive professional and customer data makes third-party dependency security a material operational risk.

The Wall Street Journal reported at least seven class-action lawsuits following the breach, including allegations about applicant data, recorded interviews, and worker monitoring.

Demand from AI labs and enterprises for specialized training data supports Mercor's expansion beyond recruiting. TechCrunch reported that the company crossed $2 billion in gross annualized revenue as of June 2026, citing its chief executive; this is a run-rate measure rather than audited full-year revenue.

The same market is supporting rapidly growing suppliers including Handshake and Micro1, indicating substantial demand alongside expanding competition. Mercor's enterprise deployment and data-licensing activities broaden its opportunity, but future growth still depends on customer spending and the usefulness of its expert-generated data.

Mercor sells expert-produced training data to frontier labs: WIRED identified OpenAI and Anthropic in April 2026, while August reporting described NVIDIA purchases for Nemotron. OpenAI said its existing projects continued after the March breach; Meta paused all work indefinitely in April. Ramp and Cognition are named benchmark collaborators, but public pages do not establish paid contracts.

Reported gross annualized revenue crossed $2 billion in June 2026, a measure of marketplace activity that does not establish recurring net revenue. Mercor sells evaluations through expert staffing or task-based delivery, yet public sources do not disclose customer counts, retention, contract commitments or account revenue shares. Its June statement that all frontier labs increased work is unquantified and does not specifically confirm Meta resumed purchasing.

Mercor sells evaluation work through expert staffing and managed task-based delivery, with validated evaluations priced per task rather than per person. Customers contact its sales team for scoped engagements and retain the resulting reusable evaluation suite; the company advertises delivery within six weeks.

For business-data licensing, Mercor pays contributing companies when anonymized datasets reach buyers, typically two to four weeks after signing, and covers extraction and processing costs.

Mercor connects professional expertise with AI development through human training data, model evaluations, and benchmarks built around economically valuable work. Its network includes physicians, lawyers, engineers, and consultants serving frontier-model and enterprise use cases.

The company also helps enterprises deploy production AI with embedded engineers and turn operational workflow data into licensed datasets. This combines expert labor, evaluation infrastructure, and enterprise implementation in a broader business than its original recruiting focus.