AI litigation in the US has expanded far beyond copyright disputes. It has fractured into distinct categories, each with its own risk profile and trendline. And the target is shifting. The first wave of litigation targeted model developers. The current wave increasingly centers on deployers, the organizations using AI tools in business operations, workflows, hiring pipelines, and customer-facing applications. Below is a survey of the major categories and trendlines.
1. Copyright and Training Data
This remains the largest category by volume and dollar exposure, and 2025 and 2026 produced the first real doctrine. Two decisions frame the issues. In Thomson Reuters v. Ross Intelligence, No. 1:20-cv-00613 (D. Del.), the court rejected a fair use defense where a legal research startup trained on Westlaw headnotes to build a competing product, emphasizing market substitution. In Bartz v. Anthropic, No. 3:24-cv-05417 (N.D. Cal.), Judge Alsup held in June 2025 that training on lawfully acquired books was “quintessentially transformative” fair use, but that downloading and retaining pirated copies was a separate act of infringement not immunized by downstream training use. That distinction drove Anthropic’s $1.5 billion class settlement, the largest copyright settlement in US history (covering roughly 500,000 works at approximately $3,000 per work). The court granted final approval of the settlement on July 20, 2026, overruling all objections; roughly 350 authors opted out, some to pursue individual claims. Kadrey v. Meta, No. 3:23-cv-03417 (N.D. Cal.), reached a similar fair use result on training while leaving piracy-based theories alive. The plaintiffs there are seeking Ninth Circuit review of the piracy holding. Notably, no federal appellate court has yet ruled on whether AI training constitutes fair use.
The flagship case remains The New York Times Co. v. Microsoft Corp. and OpenAI, No. 1:23-cv-11195 (S.D.N.Y.), now the subject of consolidated multidistrict litigation. The court largely denied the motion to dismiss in April 2025. On August 6, 2026, the court dismissed The Times's contributory infringement and trademark dilution claims, leaving direct copyright infringement, vicarious infringement, and DMCA copyright management information removal as the surviving theories. The Times filed its Third Amended Complaint on August 21, 2026, teeing up perhaps the most consequential ruling yet on whether AI training and output lawfully substitute for the market for copyrighted journalism.
The trend: Three developments stand out in AI copyright litigation. First, courts are converging on an approximate rule: lawfully sourced training data supports a strong fair use defense, while pirated source material may give rise to independent statutory damages exposure. Second, the fight is migrating from inputs to outputs. Disney v. Midjourney (C.D. Cal., filed 2025) and similar suits test liability for infringing generated content rather than the training process itself. The visual-arts cases against Stability AI, including Getty Images’ N.D. Cal. suit, and the Andersen artist class action, press the same output-side theories. Third, litigation is driving licensing deals for training content. Licensing regimes were promoted in the March 2026 White House National Policy Framework for AI as a means to protect creators: “Congress should consider enabling licensing frameworks or collective rights systems for rights holders to collectively negotiate compensation from AI providers, without incurring antitrust liability.” Major content owners are increasingly signing AI licensing deals, and settlements like Bartz may establish a baseline valuation for those negotiations. The music industry illustrates the dynamic: the record labels’ infringement suits against Suno and Udio have already yielded settlements paired with licensing arrangements.
2. AI Product Liability and Safety
The fastest-emerging category tests the bounds of First Amendment protection for AI system outputs, as well as whether AI systems, particularly chatbots, can be treated as defective products. In Garcia v. Character Technologies, No. 6:24-cv-01903 (M.D. Fla.), the court in May 2025 allowed wrongful death and product liability claims to proceed against a chatbot developer following a user’s suicide. The court, at the pleading stage, rejected the argument that chatbot outputs were categorically protected speech. Since January 2026, Character.AI and Google have engaged in multiple settlements tied suits involving user harm.
Raine v. OpenAI, No. CGC-25-628528 (Cal. Super. Ct., S.F. Cnty.), filed in August 2025 after a suicide, is another wrongful death, product liability, and negligence case against a frontier model developer and its chief executive personally. It pleads strict product liability and negligence, and an amended complaint added intentional misconduct allegations seeking punitive damages. In February 2026, California coordinated roughly a dozen related suits into a single proceeding (the state-court equivalent of MDL), In re ChatGPT Product Liability Cases, JCCP No. 5431. On June 1, 2026, Florida’s Attorney General filed a first-of-its-kind state enforcement action against OpenAI on similar deceptive trade practices and safety grounds.
The trend: AI speech and product liability cases indicate that plaintiffs are increasingly framing AI as a product in an effort to invoke strict liability, design defect, and failure-to-warn doctrines. Early court rulings have permitted those theories at the pleading stage. Dispositive rulings expected in late 2026 will shape theories of exposure for companies that ship consumer-facing AI features. State legislatures are reinforcing the trend, with California’s SB 243 companion chatbot law, effective January 1, 2026, and dozens of similar bills pending and enacted nationwide.
3. Algorithmic Discrimination and Employment
The defining case in this area was Mobley v. Workday, No. 3:23-cv-00770 (N.D. Cal.), in which the plaintiff alleged that AI screening tools discriminated on the basis of race, age, and disability. The court allowed claims to proceed past the pleading stage on the theory that a software vendor can act as an “agent” of its customer-employers, and thus the vendor can itself be liable under Title VII, the ADEA, and the ADA (a position the EEOC supported in an amicus brief). In May 2025, the court preliminarily certified a nationwide ADEA collective that could encompass millions of applicants screened through Workday’s tools, converting a novel pleading theory into large-scale discovery.
The trend: Both employers and vendors have risk exposure. Neither the employer's defense ("the algorithm did it") nor the vendor's ("we just make the software") forecloses algorithmic discrimination liability, and neither can escape liability simply by pointing to the other. Further, new statutory theories beyond discrimination are taking shape. A January 2026 class action against Eightfold AI alleges that an AI hiring platform operates as a consumer reporting agency subject to the Fair Credit Reporting Act’s disclosure and consent requirements. Under this theory, even a bias-free tool can generate liability over how it collects and scores applicant data.
4. AI Defamation
The defamation-by-hallucination doctrine is unsettled. In Walters v. OpenAI, No. 23-A-04860-2 (Ga. Super. Ct., Gwinnett Cnty.), the court granted summary judgment in favor of OpenAI in May 2025 in one of the first decided AI defamation cases. The court concluded that a ChatGPT output could not reasonably be understood, under the facts presented, as stating actual facts, when considering the context and accuracy disclaimers. The plaintiff thus could not establish negligence or actual malice given OpenAI’s established hallucination-reduction efforts. The holding there should be construed as narrow, however, because the recipient of the output knew the output was fabricated and never republished it. Other pending cases test a different set of facts. Starbuck v. Google, No. N25C-10-211 (Del. Super. Ct., filed October 2025), alleges that Google’s LLMs generated false accusations including claims of sexual abuse. LTL LED, LLC v. Google, No. 25-cv-02394 (D. Minn., filed June 2025), involves a Google “AI overview” that allegedly informed the public that a business was being sued by a state attorney general. The emerging trend: accuracy disclaimers, documented hallucination-mitigation efforts, and a defendant’s response to notice of false outputs are now material facts in AI defamation cases.
5. Privacy and Chatbot Wiretapping
This category primarily targets AI deployers rather than developers. Allegations center on website chatbot and session-tool interception, recording, or sharing of visitor communications with third-party vendors. They also raise the question whether the use of such communications for AI training without all-party consent violates the California Invasion of Privacy Act (CIPA) and the federal Wiretap Act. California’s SB 690, which originally proposed a broad commercial-purpose exception to CIPA, stalled in 2025 but was revived and narrowed in July 2026; as amended, the bill would strip the private right of action for website-based pen register claims (leaving CIPA’s wiretapping provisions untouched) and faces an August 31, 2026 deadline for floor passage. Until reform arrives, a patchwork of inconsistent rulings remains for businesses to navigate. Newer suits based on alleged transmission of private user data to third parties for targeted advertising add consumer protection claims based on misrepresentations of product privacy.
6. AI Regulation Challenges
Litigation is also emerging over AI laws rather than AI systems. In 2026, xAI asserted a constitutional challenge to Colorado’s AI Act (SB 24-205), which would require developers and deployers of high-risk AI systems to exercise reasonable care to protect consumers from known or foreseeable risks of algorithmic discrimination. On April 24, 2026, the DOJ moved to intervene in support of xAI, and enforcement of the statute was suspended by agreement a few days later pending the litigation. Colorado’s legislature responded in May 2026 by advancing a narrower framework replacing SB 24-205’s comprehensive risk-management regime with a narrower notice-and-disclosure framework limited to automated decision-making used in consequential decisions. First Amendment, dormant commerce clause, and preemption challenges will likely increase as a standalone category as states continue to enact AI laws and regulations. Compliance obligations will remain a moving target while those challenges play out.
7. Voice, Likeness, and Digital Replicas
A further category is growing quickly: AI cloning of voices and likenesses. In Lehrman v. Lovo (S.D.N.Y.), voice actors allege that an AI voice startup cloned their voices without consent; in 2025, the court allowed contract and New York right-of-publicity claims to proceed while dismissing most federal theories. Layered on top are newer state digital replica statutes, such as Tennessee’s ELVIS Act and California’s expanded publicity protections, and the pending federal NO FAKES Act. Organizations using synthetic voices or likenesses in marketing, customer service, or training content should treat consent and provenance documentation as a present-day compliance requirement rather than a future one.
8. AI-Washing and Securities Litigation
AI exposure has also reached the securities docket. The SEC has brought enforcement actions against investment advisers and public companies for overstating AI capabilities, a practice regulators call “AI washing,” and shareholder class actions asserting similar theories continue to accumulate against public companies whose AI-related statements allegedly outran reality. The trend: for public companies, statements about AI capabilities, adoption, and risk are disclosure issues that warrant the same substantiation discipline as any other material claim, and disclosure committees should review AI-related statements in filings, earnings calls, and marketing materials.
What these AI litigation trends mean for organizations deploying AI
Your risk profile is expanding. The focus in AI litigation is shifting from model developers to deployers. Chatbot wiretap claims, hiring discrimination claims, and consumer protection claims are increasingly targeting the company that deployed and used the AI tool, not just the company that built it.
Know your training data and your vendors’ training data. The emerging copyright rule rewards lawful sourcing. Vendor transparency about data provenance is key, and so is carefully negotiated indemnification for IP claims arising from AI outputs.
Treat consumer-facing AI as a product. Courts are increasingly allowing strict liability and failure-to-warn theories to proceed past the pleading stage. Document safety testing, guardrails, intervention protocols, and the reasoning behind design choices.
Audit AI in hiring. Neither employers nor the vendor tools they rely on are insulated. Document bias testing and meaningful human-in-the-loop mitigations. Track evolving state AI hiring laws.
Build a record. Across every category, documented mitigation efforts (hallucination reduction, bias audits, safety protocols, accurate disclosures) are central to the defense.
For questions about AI governance, document integrity, and litigation technology risk, please contact the Jones Walker Privacy, Data Strategy and Artificial Intelligence team. Stay tuned and subscribe for continued insights from the AI Law and Policy Navigator.
The target is shifting. The first wave of litigation targeted model developers. The current wave increasingly centers on deployers, the organizations using AI tools...