Scale AI has filed suit against a former employee and against Mercor, a fast-rising rival in AI data operations, alleging an attempt to siphon off some of Scale’s largest enterprise customers through misuse of confidential materials. The complaint accuses Mercor of trade secret misappropriation and asserts the ex-staffer, Eugene Ling, breached contractual obligations while still on Scale’s payroll.
Allegations center on “Customer A” and strategic files
At the core of the filing is an unnamed “Customer A,” which the complaint portrays as one of Scale’s biggest accounts. Scale says Ling began introducing Mercor as a potential vendor to this customer before his departure, and that he retained internal documents detailing customer strategies, pricing, work plans, and delivery playbooks. The company claims those materials would give Mercor a blueprint to serve not just Customer A, but several other marquee clients—contracts that Scale says could be worth millions of dollars to a competitor.
Trade secrets cases in enterprise data services often hinge on whether the disputed materials qualify as confidential competitive assets and whether they were used for commercial gain. Under the Defend Trade Secrets Act and state analogs, courts typically weigh the specificity of the alleged trade secrets, the security measures used to protect them, and the timing of any competing solicitations.
Mercor and the former employee push back
Mercor co-founder Surya Midha has denied that his company used Scale’s data, while acknowledging that Ling may have retained some files after leaving. He said Mercor offered to cooperate, including arranging deletion or another resolution, and was awaiting direction from Scale. Ling, in separate comments, said he never used the materials in his new role, described the retention as inadvertent, and indicated he was waiting for guidance to resolve the issue.
These responses are likely to influence any early court decisions on injunctive relief. Judges frequently look at whether the accused party moved quickly to preserve, quarantine, and return disputed files and whether there are credible explanations for their possession.
Why this fight matters in the AI vendor ecosystem
The stakes are high because large AI data contracts are unusually concentrated and sticky. Public disclosures from Appen, one of the few listed data-annotation firms, show how a single hyperscale customer can represent a double-digit share of annual revenue—making any vendor switch consequential. More broadly, Gartner and IDC have noted that enterprise AI spend remains clustered around a small cadre of technology buyers, which raises the value of each flagship account for service providers like Scale and Mercor.
The competitive backdrop has intensified as foundation models proliferate and enterprises demand higher-quality, domain-specific data. Mercor has positioned itself as a premium provider by tapping subject-matter experts—often PhDs—to curate and evaluate training data for large language models. That approach contrasts with traditional crowdsourcing and speaks to the growing emphasis on accuracy, safety, and evaluation metrics highlighted in the Stanford AI Index reports.
Perceived conflicts after high-profile partnerships
Industry reporting in recent months has spotlighted concerns about vendor neutrality in the AI stack. Media accounts have indicated that Scale’s deepening ties with a major platform company led some large data customers—who compete with that platform—to reassess their relationships. Even the perception of a conflict can push enterprises to run parallel pilots or rebid work, elevating competitive pressure from challengers such as Mercor, Labelbox, Appen, and Sama.
That context helps explain why a single customer dispute can ripple across the market. If courts validate Scale’s claims, rivals could face tighter scrutiny over employee onboarding and data hygiene. If the claims falter, it may embolden customers to diversify vendors more aggressively without fear of legal crossfire.
What courts will look for next
Early phases typically focus on temporary restraining orders, forensic reviews of devices and cloud accounts, and preservation of evidence. Subpoenas to “Customer A” could clarify whether any Scale materials were used to shape competing proposals or delivery plans. Courts also often parse the scope and enforceability of non-solicitation and confidentiality clauses, especially in jurisdictions like California that restrict non-competes but uphold trade secret protections.
For customers, the practical takeaway is straightforward: insist on clean-room practices, documented handoffs, and hard boundaries between competing vendors. For AI data providers, the moment underscores the need for strict access controls, automated offboarding, and auditable retention policies—areas many organizations say they have strengthened as model deployment scales.
Bottom line
Scale AI’s lawsuit frames a high-stakes clash over who controls the playbooks for serving the largest AI buyers. Mercor and the former employee deny misuse and signal a willingness to remediate. The outcome will shape not only the two companies’ relationship to a key set of accounts, but also the guardrails around hiring and competition in the lucrative, consolidating market for AI training data and evaluation services.