Is Scrimdata legit? What it licenses and what to ask before signing

Scrimdata says it licenses an anonymized copy of a company’s connected operating history, from email and chat to tickets, CRM and code, and builds AI training environments from it for AI labs, paying the company a price agreed in each contract, as of October 2026 (partner page). It publishes no amounts, and its website, registered in August 2026, names no one who runs it.

Who is behind Scrimdata?

Its website does not say. The scrimdata.com domain was registered on August 20, 2026, and its terms of service, updated the next day, name the business as “Scrim Data” and apply Delaware law, with no address. When Data Licensing Report checked on October 5, 2026, the home, partner, contact, careers and blog pages named no founders, staff or investors. Its careers page describes a “Small team” in which every role is remote.

It also names no data partners, labs or completed licenses. That leaves the company, rather than its record, as the thing to check: ask for the legal entity’s full name and registration, the people who would sign, and a reference from a company that has licensed data to it. Is it a scam? lists the checks that apply to any provider.

Is Scrimdata only for companies that are shutting down?

No. Its partner page is headed “Operating, pivoting, or winding down,” and describes three cases: an operating company licensing its history while it keeps trading, a wind-down, where that history “may be the most liquid asset you have left,” and an exit or pivot that leaves legacy systems behind. Its home page puts it as “Operating, winding down, or exited.” No size, age or country rule is published.

What does Scrimdata pay, and how is the price set?

It does not say. Asked “How much can we earn?”, its FAQ answers: “It depends on breadth, depth, and interconnection. Contact us to discuss the scope and valuation process.” The price is agreed at the contract step, alongside license scope, permitted use, retention and deletion. Nothing on the site says whether payment is a single fee or tied to Scrimdata’s sales to labs, or when it is made.

The first estimate needs no data. Scrimdata says estimates use “an inventory of tool types, team size, and years of history,” then scoping happens under NDA on “the tool stack and structural metadata, never content.” A data inventory prepared in advance answers those questions. Because Scrimdata stresses that “Interconnection matters more than volume,” it helps to show how your systems link: tickets that reference email threads, or commits tied to project tasks.

What would Scrimdata do with our records?

Build training environments for AI labs from an anonymized copy. Its home page describes environments of multi-step, multi-tool tasks, expert rubrics and programmatic verifiers, built for AI labs training frontier models. Its partner page calls the deal “A license, not a sale”: Scrimdata receives “a written, scoped license to use an anonymized derivative for AI training,” and the page states, “You own the data before, during, and after.” The license is “Non-exclusive by default. Exclusivity is your choice.”

Scrimdata says it anonymizes the data itself, after extraction, and that extraction is automated and encrypted, using credentials “scoped per extraction session.” Its home page says its system, Curtain, transforms “Every name, person, client, dollar figure, product detail, and date” consistently, so the links between records survive, and that “The raw archive stays sealed.” Its partner page also lists “No manual review of partner content,” “No lab access to raw data,” logged access, and “Region-locked EU and US processing.” Ask what access the extraction needs to each system, how long the raw archive is kept, and who can unseal it. What “anonymized” means explains why consistent substitution differs from deleting names.

Could Scrimdata name our company publicly?

Yes, unless you tell it not to. Its terms of service say: “Unless you tell us otherwise in writing, if your organization becomes a customer or data partner we may identify it by name and logo in customer lists and marketing materials.” The same terms say a signed data license agreement governs the datasets themselves. If you want the relationship kept confidential, say so in writing and put it in the license. What’s the catch? covers how other providers handle being named.

What should we get in writing from Scrimdata?

The signing company’s full legal name, registration and officers come first, since the site names none. The license should then say whether payment is a single fee or tied to Scrimdata’s sales, and when it is made; what access extraction needs and for how long; how long the raw archive is kept, who can open it and how deletion is confirmed; which labs receive environments built from your records; and that Scrimdata will not name you in customer lists or marketing. Ask, too, to speak with a company that has already licensed data to it.