Where these cases are being filed
AI litigation in the United States has clustered around a handful of plaintiff groups: rights holders challenging how models were trained, people alleging privacy or biometric violations, applicants and consumers challenging automated decisions, and businesses in contract disputes with AI vendors. Many of these cases are filed as class actions, and several have been consolidated in federal court, including in New York, so early rulings on motions to dismiss and class certification carry weight. Courts are still deciding how existing doctrines such as fair use apply to AI systems, and the results so far have not pointed in one direction. A company can also face parallel attention from regulators or securities plaintiffs over what it said publicly about its AI.
Discovery when the evidence is a model
Ordinary discovery rules apply, but the material is unusual. The other side may seek training data, model weights, source code, prompt and output logs, and internal evaluation results, much of which is commercially sensitive or subject to licenses with third parties. Courts typically manage this through protective orders, inspection protocols, and limits on who can see what, and those terms are negotiated early and contested often. Preservation is a frequent battleground, because logs and intermediate model versions are routinely overwritten and courts have ordered parties to retain data they would normally delete. Put a litigation hold in place promptly, and involve engineers who understand where the data actually lives.
Indemnities, insurance, and case planning
Before strategy is set, look at who else may share the cost. Some AI vendors offer indemnities for certain claims about outputs, usually with conditions the customer must have met, and enterprise contracts often allocate infringement and data risks in detail. Insurance coverage for AI claims is uneven, so policies should be read closely and insurers notified promptly. Our early work maps your role as developer, deployer, or user against the claims pleaded, identifies motions that could narrow the case, and plans how technical witnesses will explain the system. We are direct about which legal questions remain open, since that uncertainty shapes settlement discussions as much as the facts do.
