Riding AI litigation's second wave with AI

The first AI litigation wave focused on copyright. As this wave crests, with early indicators from Judges Alsup, Chhabria, Bibas, Stein, among others, a second swells. The next wave will shift from training LLMs to deploying agents. These agents, in the virtual and physical worlds, present use cases that may upend every form of work--including that of lawyers. Agent proliferation implicates novel issues around the Computer Fraud and Abuse Act, online contract, trespass, tortious interference, unfair competition, antitrust and product liability.

Given the accelerating maturation of AI tooling, lawyers litigating this agent-focused set of cases will have powerful tools available. This article reviews the coming second wave of agent-focused AI cases and how lawyers can use AI when litigating them.

Agentic use cases. Agents are first born in the computer (as Zoolander predicted). A common first use case is web browser based. These agents search the web and navigate it on behalf of humans. Armed with human user credentials, these browser-based agents can find information, make purchases and post content. API-centric agents go a step further. These work programmatically with exposed endpoints across platforms. They read information, write data and develop new software with API building blocks. One can imagine a world where user interfaces are streamed on the fly, rather than static, built by agents working programmatically to perform software tasks.

Embodied agents are next. Today, agents must resort to asking humans to do things for them in the physical world (as the tagline of rentahuman.ai explains: AI needs your body). Soon, LLMs will exist in robots. Agents in the material world will be able to walk, carry, and interact with physical objects, other agents and humans.

Litigation claims. The last decade's scraping battles foreshadowed the next decade's bot fights.

• CFAA. As agents proliferate, platforms will invoke the Supreme Court's Van Buren v. United States "gates-up-or-down" approach when they exclude them. Agent developers will attempt to defend themselves through the human's permission to use their credentials. This will raise classic issues around how informed the consent is and whether handing over passwords to agents works from a CFAA perspective. Trespass causes of action will face similar dynamics.

• DMCA. Copyright law will flow through the second wave to the extent that agents use their advanced coding prowess to circumvent technical measures aimed at blocking software access. The specter of DMCA statutory damages raises the stakes--as we saw in the DRM battles of yesteryear.

• Tortious interference, unfair competition and antitrust. Many agent cases will revolve around upstarts and competitors attempting to siphon information from incumbent systems of record. In some cases, those agents will attempt to inject information into the systems too. This will occur against the backdrop of go-to-market messaging and customer communications. Circumstances ripe for claims of interference with contract, prospective economic relationships and false advertising. An evolved version of the antitrust battles around operating systems, browsers and house apps is likely to surface in the second AI litigation wave.

• Product liability. Physical agents embodied in robots will implicate the same claims as software agents, because they likely will be Internet connected. Physical agentic robots will also spark classic product liability disputes, to the extent they cause harm to property or people. There is a wide spectrum, ranging from the cleaning bot jostling a stack of valuables to... Skynet. Today, thousands of cases across the country are grappling with the extent to which online platforms can be considered products for purposes of liability. When the platform is walking around stacking boxes, it may not be so close a call.

AI tooling for litigating AI waves. Opinions about AI tooling are like opinions, everybody's got one. By some counts, over 500 rulings have resulted from unfortunate situations of AI hallucinations creeping into briefs (and even judicial decisions). Using AI tools for litigation is no doubt high-risk. But this technology cannot be ignored. The introduction of powerful reasoning models changed the game last spring. Over the last few months, agents have changed it again. A few practical suggestions for litigators wading into these waters:

• No writing. Some day the LLMs might write better than a well-trained litigator. That day is not today. Removing writing from AI litigation workflows can salve a lot of pain. If text is never copied from an LLM to a draft, but instead the lawyer drafts every word, this reduces risk of hallucination. It also maintains style, flow and personality.

Call it AI psychosis if you will. But short of writing, AI tools--with today's capabilities--can revolutionize every other aspect of litigation in ways we could not have imagined a year ago.

• Outline. The most powerful reasoning models, today, can crush outlines. Give them prior briefs as context and key legal authorities and they can spin outlines in minutes that would take a strong team of litigators 50+ hours.

• Audit and revision. Top-level models are fantastic at reviewing prior briefs and noting what arguments went unaddressed and what merits emphasis. They can formulate strong revision plans based on a few comments from a human editor. No brief should be submitted without running through an AI proofer. At 3 in the morning, before that draft goes to the client, nothing beats a robot hunting typos.

• Don't sleep on trial applications. Feed a model transcripts, get back gold. AI can take a chart of evidence objections and flow it through hundreds or thousands of exhibits. Spotting a single cell in a ten-thousand row spreadsheet that impeaches the witness--easy work for AI agents. Particularly in Temporary Restraining Order and Preliminary Injunction trials, where much of AI litigation battles are fought and records are underdeveloped and time is short, AI is invaluable.

• Train and scale. It is critical that partners train their teams on best practices. Do not copy paste case citations. Do not copy paste anything. Human-verify it all (Battlestar Galactica showed us the way). Teams should share AI chats just as they might share drafts and redlines. Maintain version control. Agents can help--LLMs are excellent at formulating feedback.

We may not know exactly where this all leads. But we know that we are in for a flood of access-based cases due to AI agents. Litigators riding this wave will do well to follow the engineers. Where coding tools were a couple years ago, emerging, AI use cases for lawyers similarly are emerging today. Many top tier engineers will report that their workflows in 2026 are unrecognizable from those in 2024. Litigators are in for the same. When this agent-litigation wave recedes and the next begins to form, perhaps we will be watching our agentic colleagues colloquy with robot judges. Whatever happens, we are in for a fun surf session.

Michael Burshteyn is a shareholder at Greenberg Traurig, LLP.