Products
Our proprietary Croquis tool detects creative works inside AI training datasets and models. Put it to work for your law firm or litigation team, at the scale your cases demand.
Courts are deciding, case by case, how copyright law applies to generative AI. Millions of dollars and your client’s IP are at stake while the evidence you need sits inside terabyte-scale datasets and the models themselves—where traditional discovery often can’t reach. But our custom detection technology can.
Answer: let your clients know if their IP is being used without their knowledge or consent.
Assess: find out whether a claim has a factual footing before you pursue it—or screen a portfolio to find the claims worth bringing.
Prove: once litigation is underway, build documented, reproducible evidence for the specific works at issue.
See sample Croquis findings here.
Whether you’re evaluating a potential claim, screening a portfolio of prospective plaintiffs, or deep in discovery—we’d like to hear about the case.
Establish the factual basis for plausible allegations before you file—or find the cases worth bringing.
Turn a well-founded suspicion into documented, reproducible evidence—work by work.
Factual contentions in a complaint need evidentiary support—or a reasonable basis to expect it after discovery. Dataset and model analysis turns “we believe our client’s work was used” into specific, checkable findings you can stand behind at filing.
Knowing which datasets, model versions, and behaviors matter lets you draft discovery that asks for the right things—and independent technical analysis gives you a benchmark to evaluate what the other side produces.
When claims span thousands of works or many rights holders, the analysis has to scale with them. Per-work findings across a whole portfolio help show the common patterns that portfolio-wide and class claims are built on.
We provide technical analysis, not legal advice. Our reports and methods are built to support your case team and your experts—the legal strategy stays yours.
The works, the rights holders, and the models or datasets in question—whether you’re screening a potential claim or already in active litigation. Handled confidentially.
We search public training datasets and probe the models across the works at issue, with methodology documented and every result logged for reproduction.
Per-work reports with supporting outputs and reproducible protocols—ready for your case team and experts to assess, cite, and act on.
