idler data licensing review: what it builds from your records and what to ask

idler, a San Francisco AI data lab from Y Combinator’s Summer 2025 batch, invites companies with 10 to 100 employees to “supplement your revenue by licensing your codebase or de-identified business data,” and builds evaluations and training environments for AI labs from partners’ work (YC profile, as of October 2026). It publishes no payment amounts, structure or timing. idler says two of its collections were built from real companies’ records: an e-commerce company’s and a law firm’s.

Who would we license to, and what does idler do with the data?

You would license to idler itself. Its YC profile says “idler is a frontier data research lab” and “We build the evals and environments that the world’s leading frontier labs use to measure and train their models.” It names none of those labs. The invitation to companies is on that profile, not on idler.ai, whose “Contact” link opens an Airtable form; the profile asks companies that fit to write to hello@idler.ai. Its range starts at 10 employees and stops at 100; most floors other providers publish start at about 20, as Does my company’s data qualify? shows.

When Data Licensing Report checked on October 5, 2026, idler.ai had no terms of service or privacy policy and named no legal entity or address. The license agreement would be the first document to say who signs, what idler and the labs may do with your data, and for how long.

What did idler build from a law firm’s data?

A private legal benchmark called CorpLaw, which idler describes as “A legal benchmark built from a real law firm’s anonymized data”. Its 50 tasks put an AI agent into “a live transaction workspace containing a data room of document versions, negotiation email threads, defined terms, and client instructions, all drawn from the firm record.” Grading follows criteria “grounded in the work the firm produced on each matter, then extended and pruned through attorney review at the firm,” so lawyers at the firm worked on the benchmark as well as supplying records. The page marks the dataset private, and the collections page labels it “Access Upon Request.”

The CorpLaw page names no firm and does not say how client consent or privilege was handled, whether the firm was paid, or on what terms (checked October 5, 2026). Negotiation emails and client instructions are information relating to a client’s matter, the kind ABA Model Rule 1.6 protects. Can a law firm sell its data to AI companies? sets out that rule, privilege and the questions a firm would take to its own lawyer before any matter file goes out.

Could our records end up in a public benchmark?

They could. idler has already published results built from one partner’s records, describing the partner without naming it. idler calls ShelfLife “our first public benchmark,” a “digital twin seeded with the complete operational data of a live, profitable, growing multi-brand retailer running 9 stores, 3 commerce platforms, and a multi-supplier sourcing operation” (YC profile). The ShelfLife page publishes scores for six AI models on 200 environments “drawn from the company’s operational data over the past decade,” and offers the full dataset on request.

idler names neither source company but describes both: the retailer by its stores, platforms and sourcing, the law firm by its corporate transaction work. Ask whether a collection built from your records would be public or private, how your company would be described in it, and who may request the full dataset.

What does idler pay, and when?

idler does not say. Its one statement about compensation is the invitation to “supplement your revenue by licensing,” and neither the YC profile nor idler.ai’s home, collections and careers pages give an amount, a structure or a payment date (checked October 5, 2026). The ShelfLife and CorpLaw pages say nothing about what either source company received.

With no published figure, the first number you see will be idler’s own offer. How offers work covers getting a second offer to compare it with, and How long until I get paid? lists the timing terms to fix in writing.

How much of our team’s time would it take?

More than an export, going by the two partnerships idler describes. For ShelfLife it says “Every environment is constructed in direct collaboration with the company’s leaders and the operators who performed the work in production,” and that it “worked iteratively with them” (ShelfLife). CorpLaw’s grading criteria went through attorney review at the firm. A Forward Deployed Engineer posting describes engineers who “own the entire system that turns our partners’ real work into evals & RL environments” and asks for willingness to travel.

Ask which of your people idler would need, for how many hours, and whether that time is paid. How much work is required? covers the export side.

How established is idler?

It is a 2025 startup with venture funding. Its YC profile lists it as founded in 2025, in the Summer 2025 batch, with 12 people, and names three founders: Ivan Chub (CEO), Nalu Concepcion and Tony Goss (CTO). RuntimeWire reported in August 2026 a $9 million seed round led by Paradigm, with Y Combinator, Long Journey and angel investors, and FinSMEs reported the same round. The coverage Data Licensing Report found on October 5, 2026 was funding news and directory listings.

Its record with data partners is its own account: the two collections above, and its statement that post-training Nemotron 3 Nano on 41 ShelfLife tasks raised the model’s Finance Agent Benchmark score “from 30.0% to 41.8%.” Neither source company is named, so a reference from one of them would have to come through idler.

What should we settle with idler before licensing?

idler’s site names no legal entity and publishes no payment terms, so those come first, followed by the license itself and what happens to your records:

1. Which legal entity signs, and whether we can read the agreement before sharing any sample.
2. What idler pays, in what structure, and on what date.
3. Whether the license is exclusive, how long it lasts, and what idler and the labs may do with the data.
4. Whether anything built from our records will be public, and how our company will be described.
5. Who removes identifying details, before or after transfer, and who at idler sees the raw records.

Get the answers in writing, along with idler’s estimate of the staff time it needs from you.