AI lawsuits explained: who is getting sued?

Case status as of October 2026. Statuses change frequently; each entry reflects the latest ruling or filing confirmed at that date.

AI lawsuits are claims that AI developers, or the organizations deploying AI, infringed rights or caused harm. Most allege copyright infringement in training data or outputs, but a fast-growing share concern chatbot harm, discriminatory decisions, privacy and misleading AI-generated statements. The defendants have broadened too: model developers were the first wave, and deployers, software vendors and even the lawyers who rely on AI are now in court.

Courts increasingly treat three questions separately: whether training a model is lawful, how the training data was obtained, and whether the model's outputs infringe. That separation produced the largest copyright settlement in US history.

Status table: major AI cases as of October 2026

Case	Court	Issue	Status (October 2026)	Source
Bartz v. Anthropic	N.D. Cal.	Training on books; pirated library	Settled, US$1.5B; final approval 20 Jul 2026; fee-award appeals only	Final approval order
Kadrey v. Meta	N.D. Cal.	Training Llama on books; torrenting	Fair use for Meta on training (Jun 2025); distribution (torrent upload) claim pending, summary judgment hearing set for 25 Feb 2027	Order, Mar 2026
Thomson Reuters v. ROSS Intelligence	3d Cir.	Headnotes used to train AI legal research	Affirmed for Thomson Reuters, 29 Sep 2026: not fair use	Docket
In re OpenAI Copyright Infringement Litig. (NYT, Daily News, Authors Guild and others v. OpenAI, Microsoft)	S.D.N.Y. (MDL 3143)	Training and outputs; news and books	Summary judgment briefing (Sep–Nov 2026); DOJ statement of interest backing fair use (1 Sep 2026)	MDL docket
Getty Images v. Stability AI (UK)	High Court (Ch); Court of Appeal	Secondary infringement; trade marks	Getty largely lost (4 Nov 2025); permission to appeal on "infringing copy" granted Dec 2025; appeal pending	CMS case tracker
Getty Images v. Stability AI (US)	N.D. Cal.	Training on 12M+ images; watermarks	Motion to dismiss largely denied (Apr 2026); proceeding	Docket
Disney, Universal & Warner Bros. v. Midjourney	C.D. Cal. (consolidated)	Training and character outputs	Discovery; consolidated Nov 2025	Docket coverage
Concord/UMG/ABKCO v. Anthropic (I and II)	N.D. Cal.	Lyrics; torrented songbooks	Case I in discovery (preliminary injunction denied Mar 2025); Case II filed 28 Jan 2026 (US$3B+ claimed)	MBW
Sony Music Publishing & Warner Chappell v. Anthropic	N.D. Cal.	Compositions; founders named	Filed 28 Aug 2026	MBW
UMG & Sony v. Suno; Sony v. Udio	D. Mass.; S.D.N.Y.	Sound recordings; stream-ripping	Warner settled with both (Nov 2025), UMG with Udio (Oct 2025); UMG and Sony filed a second Suno suit 18 Sep 2026; Sony filed a new Udio suit 20 Jul 2026	Complaint, Sep 2026
GEMA v. OpenAI	LG München I	Memorized lyrics in model and outputs	GEMA won (11 Nov 2025); OpenAI appealing	CMS
GEMA v. Suno	LG München I	Training and outputs; TDM exception	GEMA largely won (31 Jul 2026)	Simmons & Simmons
Like Company v. Google (C-250/25)	CJEU (Grand Chamber)	Is LLM training reproduction? TDM exception	Heard 10 Mar 2026; judgment awaited	CURIA
Toronto Star et al. v. OpenAI	Ont. S.C.J.	News content; jurisdiction	Jurisdiction upheld, 2025 ONSC 6217 (7 Nov 2025); proceeding	CanLII
Garcia v. Character Technologies (and four related cases)	M.D. Fla. and others	Wrongful death; product liability; minors	Settled Jan 2026 (terms confidential)	Court filing
Raine v. OpenAI	Cal. Super. Ct. (San Francisco)	Wrongful death; design defect	Pending; no trial date	NBC News
Mobley v. Workday	N.D. Cal.	AI hiring discrimination; vendor as agent	ADEA collective certified (May 2025); class certification motion filed Sep 2026, hearing 9 Mar 2027	Docket
Moffatt v. Air Canada	BC Civil Resolution Tribunal	Chatbot misrepresentation	Decided 14 Feb 2024; Air Canada liable	Decision

Model developers: copyright and training data

Books: Bartz, Kadrey and what followed

Two June 2025 rulings in the Northern District of California set the pattern.

• In Bartz v. Anthropic (23 June 2025), Judge William Alsup held that training Claude on lawfully acquired books was fair use, calling it "exceedingly transformative". He held that downloading and keeping millions of pirated books in a central library was a separate use, and not protected.
• Two days later in Kadrey v. Meta, Judge Vince Chhabria granted Meta summary judgment on fair use for training Llama. He stressed the ruling was confined to the 13 plaintiffs' failure to prove market harm, and suggested "market dilution" by AI-generated substitutes could win in a better-developed case. The plaintiffs' separate claim that Meta distributed books while torrenting them survives; summary judgment on it is set for 25 February 2027.

Facing a class trial on piracy, Anthropic settled. The US$1.5 billion Bartz settlement covers roughly 500,000 works at about US$3,000 each, four times the US$750 statutory minimum, and requires destruction of the pirated datasets. After Judge Alsup retired at the end of 2025, Judge Araceli Martínez-Olguín granted final approval on 20 July 2026, cutting class counsel's fee request from US$187.5 million to about US$101.5 million. Appeals filed in August 2026 challenge only the fee award, and Anthropic funds the settlement in instalments through 2027.

The Bartz outcome is a settlement, not a damages award. It releases past conduct only, not future training or outputs.

The piracy theory has since been pleaded widely: by textbook authors against Meta (July 2026) and OpenAI and Microsoft (August 2026), by Hachette, Cengage, Elsevier and Scott Turow against Google over Gemini (July 2026), and by music publishers against Anthropic (below).

The first appellate ruling: Thomson Reuters v. ROSS

On 29 September 2026 the Third Circuit affirmed the district court in Thomson Reuters v. ROSS Intelligence. ROSS had copied Westlaw headnotes, via a contractor, to train an AI legal research tool that competed with Westlaw. The court held:

• the headnotes are original enough to be protected;
• the use was commercial and only minimally transformative; intermediate copying can be justified by necessity, but not by convenience;
• the use harmed both Westlaw's market and a developing market for licensing content as AI training data.

The opinion was initially filed under seal and released publicly on 30 September 2026. It is the first appellate ruling on fair use in AI training, but it is narrow. ROSS was a non-generative tool that returned existing judicial text, and the court expressly did not decide how fair use applies to generative AI. Its market-harm reasoning, recognizing a licensing market for training data, is the part plaintiffs elsewhere will cite.

News publishers vs AI companies

• The New York Times v. Microsoft and OpenAI (filed December 2023) survived most of the motions to dismiss in March 2025 and was centralized with authors' and other publishers' suits in In re OpenAI Copyright Infringement Litigation (MDL 3143) before Judge Sidney Stein. Discovery fights included orders requiring OpenAI to preserve and produce millions of de-identified ChatGPT logs. In September 2026 the parties filed cross-motions for summary judgment on fair use. On 1 September 2026 the US Department of Justice filed a statement of interest arguing that LLM training is generally fair use. The news plaintiffs replied that it deserves no weight because it does not apply the fair-use factors to the facts. More publishers keep joining: The Seattle Times and Newsday (4 September 2026) and a group of regional publishers (16 September 2026).
• Perplexity faces suits from Dow Jones and the New York Post (2024; motion to dismiss denied), The New York Times (December 2025), CNN (May 2026) and others over retrieval-augmented answers that reproduce paywalled journalism.
• Canada. A coalition including Torstar, The Globe and Mail, Postmedia, The Canadian Press and CBC/Radio-Canada sued OpenAI in Ontario in November 2024, alleging copyright infringement, circumvention of technological protection measures, breach of website terms and unjust enrichment. In Toronto Star Newspapers Ltd. v. OpenAI Inc., 2025 ONSC 6217 (7 November 2025), the Superior Court rejected OpenAI's jurisdiction challenge. It held that Canadian authors of Canadian works may pursue claims under Canadian law in Canada. The merits are untested; there is no Canadian fair-dealing ruling on AI training yet.
• EU. In Like Company v. Google (C-250/25), a Hungarian publisher challenges Gemini's use of its articles. The CJEU Grand Chamber held its first hearing on generative AI and copyright on 10 March 2026. The questions are whether training is "reproduction" under the InfoSoc Directive and whether the text-and-data-mining exception covers it. A judgment is awaited and will bind all EU courts.

Images and video

• Getty Images v. Stability AI (UK). On 4 November 2025 the High Court (Joanna Smith J) largely rejected Getty's claims. Getty had dropped its primary training claim because training happened outside the UK. The court held that Stable Diffusion's model weights are not an "infringing copy" because they do not store the works, so the secondary-infringement claim failed. Getty won only limited trade mark relief over watermark-bearing outputs from early model versions. In December 2025 the judge granted Getty permission to appeal on the "infringing copy" point, a question of statutory construction with consequences for all software. The Court of Appeal has not yet ruled.
• Getty Images v. Stability AI (US). Getty refiled in the Northern District of California in August 2025. In April 2026 Judge Trina Thompson let the copyright, trade mark, dilution and unfair competition claims proceed, dismissing only the DMCA false-CMI claim with leave to amend.
• Disney, Universal and Warner Bros. Discovery v. Midjourney. Disney and Universal sued in June 2025 and Warner Bros. Discovery in September 2025. Judge John Kronstadt consolidated the cases in November 2025. The studios target both training and outputs depicting characters such as Darth Vader, Shrek and Superman. The case is in discovery; summary judgment is expected late 2026 or 2027. Separately, Evox Productions sued Stability AI, Runway, DeviantArt, Hugging Face and Midjourney in July 2026 over vehicle photographs in LAION datasets.
• Andersen v. Stability AI (artists' class action, N.D. Cal.) continues on narrowed claims.

Music

• Labels v. Suno and Udio. The majors sued both in June 2024. UMG settled with Udio in October 2025, and Warner with both Udio and Suno in November 2025, turning litigation into licensed platforms. UMG and Sony have not settled with Suno. After the court refused to add more than 60,000 works to the original case, they filed a second suit on 18 September 2026. It alleges that Suno's v6 model was trained on outputs of earlier infringing models, which the complaint calls "laundering". Sony filed a new suit against Udio on 20 July 2026 over 30,117 recordings allegedly stream-ripped from YouTube.
• Publishers v. Anthropic. In the original Concord v. Anthropic (2023), a preliminary injunction was denied in March 2025 and the case continues. UMG, Concord and ABKCO filed a second suit on 28 January 2026, claiming more than US$3 billion in statutory damages over about 20,000 works allegedly taken from torrented songbooks. Sony Music Publishing and Warner Chappell sued on 28 August 2026, naming Dario Amodei and Benjamin Mann personally.
• Germany. In GEMA v. OpenAI (Munich Regional Court I, 11 November 2025), the court held that the lyrics of nine German songs were memorized in OpenAI's models and reproduced in outputs. It rejected the text-and-data-mining defence for memorized content and enjoined the reproduction. OpenAI is appealing. In GEMA v. Suno (31 July 2026), the same chamber found six works, including "Daddy Cool" and "Forever Young", reproducibly contained in Suno's models. It found that Suno had circumvented YouTube's protection measures by stream-ripping, and ordered Suno to stop using the works, disclose revenue and pay damages.

Code

In Doe v. GitHub (Copilot), the Ninth Circuit on 16 September 2026 affirmed dismissal of the remaining DMCA §1202(b) claim. It reasoned that Copilot generated new code rather than delivering existing works stripped of their copyright management information. The decision does not address training-stage CMI removal, and the plaintiffs' licence-based contract claims remain pending in the district court.

AI search engines and agents

• Amazon v. Perplexity. Amazon obtained a preliminary injunction in March 2026 against Perplexity's Comet browser agent shopping on Amazon. On 4 August 2026 the Ninth Circuit vacated it, holding that the user, not Perplexity, accesses Amazon's systems for computer-misuse purposes. The court stressed the ruling is limited to the current record. It is the first appellate decision on agentic AI under the CFAA, and the case returns to the district court.
• Secondary liability narrowed. In Cox Communications v. Sony Music (25 March 2026), the US Supreme Court held that a service provider is not contributorily liable merely for knowing that some users infringe. Liability requires inducement or a service tailored to infringement. AI platforms are already invoking Cox against claims based on what users prompt.

Chatbot harm and product liability

• Character.AI. Garcia v. Character Technologies (M.D. Fla., filed October 2024) alleged a companion chatbot contributed to a 14-year-old's suicide. In May 2025 the court refused to dismiss negligence and product-liability claims and declined, at that stage, to treat chatbot output as protected speech. In January 2026 Character.AI, its founders and Google settled Garcia and four related family cases on confidential terms.
• OpenAI. Raine v. OpenAI (San Francisco Superior Court, August 2025) alleges ChatGPT encouraged a 16-year-old's suicide. An amended complaint alleges intentional removal of safety protocols, opening a path to punitive damages. Further wrongful-death and psychological-harm suits followed in late 2025. None has reached trial. These cases test whether a general-purpose chatbot is a "product" with a design-defect duty.
• Why it matters beyond AI labs. Duty-of-care reasoning based on "foreseeable risks associated with anthropomorphic AI systems" can reach any organization deploying a conversational agent to vulnerable users.

Hiring and discrimination

Mobley v. Workday (N.D. Cal.) is the leading AI hiring case. In 2024 Judge Rita Lin allowed claims that Workday could be liable as its customers' agent for discriminatory AI screening. In May 2025 she certified a nationwide Age Discrimination in Employment Act collective of applicants aged 40 and over, later extended to HiredScore features. Notice went out in early 2026. In September 2026 the plaintiffs moved to certify Rule 23 classes covering race, sex, age and disability; the hearing is set for 9 March 2027. The vendor-as-agent theory matters to every buyer of AI screening tools: a vendor's liability does not displace the employer's own. See bias in AI systems.

When the deployer is liable: Air Canada

In Moffatt v. Air Canada (2024 BCCRT 149, 14 February 2024), Air Canada's website chatbot told a grieving customer he could claim a bereavement fare retroactively, which contradicted the airline's actual policy. The British Columbia Civil Resolution Tribunal held Air Canada liable for negligent misrepresentation. It rejected the suggestion that the chatbot was "a separate legal entity that is responsible for its own actions" and awarded C$650.88 in damages plus interest and fees. Disclaimers elsewhere on the website did not help. See AI transparency.

Privacy and biometrics

• Clearview AI. The novel BIPA class settlement, which gave class members a 23% equity stake in Clearview instead of cash and was approved in March 2025, was vacated by the Seventh Circuit on 13 July 2026 for inadequate class representation. The case is back in the district court.
• Otter.ai. On 13 August 2026 a federal court in California let wiretap, California Invasion of Privacy Act and BIPA claims proceed. The claims allege that Otter's meeting bot recorded participants and trained on their conversations without all-party consent.
• Regulators, not just plaintiffs. Italy's Garante fined OpenAI €15 million in December 2024 over ChatGPT's legal basis and transparency (OpenAI appealed). Privacy regulators in Canada and the EU continue investigations into generative AI. For the compliance side, see AI privacy.

Lawyers sanctioned for AI hallucinations

By volume, most AI-related court decisions do not involve AI companies. Courts are sanctioning lawyers and litigants who file AI-generated briefs with invented cases.

• Mata v. Avianca (S.D.N.Y., June 2023): two lawyers and their firm fined US$5,000 for ChatGPT-fabricated citations, the case that started the trend.
• Canada: in Zhang v. Chen (2024 BCSC 285) a lawyer was ordered to pay costs personally for citing fictitious ChatGPT cases. In Ko v. Li (2025 ONSC 2766) an Ontario judge initiated contempt proceedings over a factum citing non-existent authorities. The Federal Court of Canada requires parties to declare AI-generated content in filings.
• Scale: Damien Charlotin's public database had logged more than 1,600 court decisions worldwide involving hallucinated content by mid-2026, most in the US. Sanctions have escalated from small fines to disqualification of counsel and regulatory referral.

The legal questions still open

• Is generative AI training fair use? As of October 2026, no appellate court has decided this for a generative model. The OpenAI MDL and Kadrey's next phase are the likeliest US vehicles. ROSS shows that market harm, including harm to a licensing market for training data, can defeat the defence.
• Do model weights "contain" the works? The UK High Court said no in Getty; Munich said memorized works are reproduced in the model in GEMA. The Court of Appeal and the CJEU will shape this.
• Do outputs infringe? Character, lyric and article regurgitation claims (Disney, GEMA, NYT) are stronger for plaintiffs than pure training claims.
• Who can own AI output? The US Supreme Court denied certiorari in Thaler v. Perlmutter on 2 March 2026, leaving in place the rule that copyright requires human authorship.
• What does the US Copyright Office think? Its Copyright and Artificial Intelligence, Part 3: Generative AI Training report was issued in pre-publication form on 9 May 2025. It concluded that some training uses will be fair and others will not, particularly where outputs compete with the originals or the data was pirated. A final version has not been issued. The day after the pre-publication release, the Register of Copyrights, Shira Perlmutter, was dismissed. Her challenge to the removal is still being litigated; the D.C. district court stayed it in September 2026 pending legislation.
• What does the US government argue now? The DOJ's September 2026 statement of interest backs fair use for LLM training. The Copyright Office's analysis is more cautious. Courts are not bound by either.

What this means for organizations using AI

The litigation also changes expectations for organizations that buy or deploy AI rather than develop it.

• Ask vendors about data provenance, not just capability. Bartz and the torrenting claims show acquisition is where the money is.
• Read the indemnity. Major providers offer IP indemnities for enterprise customers, with conditions such as using built-in filters. Know what voids them.
• Control outputs, not just inputs. Regurgitation of lyrics, articles or characters is the strongest claim type. Output filtering and human review for published content are practical controls.
• Own what the organization's AI says. After Moffatt, treat customer-facing AI statements as the organization's statements, and test them as such.
• Audit decision systems for disparate impact, especially in hiring, credit and insurance. Vendor assurances do not transfer liability.
• Verify before filing or publishing. The hallucination cases reflect a verification failure rather than a technology failure.
• Report it upward. Material AI litigation exposure belongs in the board pack. See AI oversight at board level and, for regulated firms, AI risk in financial services.

For the compliance frame, see the EU AI Act and AI in risk, audit and compliance. For the failure modes behind these disputes, see AI risks.

Related references

Sources and references

• Bartz v. Anthropic, final approval order (N.D. Cal., 20 July 2026): https://storage.courtlistener.com/recap/gov.uscourts.cand.434709/gov.uscourts.cand.434709.680.0_4.pdf
• Thomson Reuters v. ROSS Intelligence, docket (3d Cir. decision 29 September 2026): https://www.courtlistener.com/docket/17131648/thomson-reuters-enterprise-centre-gmbh-v-ross-intelligence-inc/
• In re OpenAI, Inc. Copyright Infringement Litigation, 1:25-md-03143 (S.D.N.Y.): https://www.courtlistener.com/docket/69879510/in-re-openai-inc-copyright-infringement-litigation/
• Getty Images v. Stability AI, 3:25-cv-06891 (N.D. Cal.): https://www.courtlistener.com/docket/71112094/getty-images-us-inc-v-stability-ai-ltd/
• Doe v. GitHub, Ninth Circuit opinion (16 September 2026): https://cdn.ca9.uscourts.gov/datastore/opinions/2026/09/16/24-7700.pdf
• UMG Recordings and Sony Music v. Suno, complaint (18 September 2026): https://www.musicbusinessworldwide.com/files/2026/09/26-cv-14275-Dkt.-1-Complaint.pdf
• Toronto Star Newspapers Ltd. v. OpenAI Inc., 2025 ONSC 6217: https://www.canlii.org/en/on/onsc/doc/2025/2025onsc6217/2025onsc6217.html
• Like Company v. Google Ireland, C-250/25 (CJEU): https://curia.europa.eu/juris/liste.jsf?num=C-250/25
• Moffatt v. Air Canada, 2024 BCCRT 149: https://decisions.civilresolutionbc.ca/crt/crtd/en/525448/1/document.do
• Mobley v. Workday, 3:23-cv-00770 (N.D. Cal.): https://www.courtlistener.com/docket/66831340/mobley-v-workday-inc/
• Garcia v. Character Technologies, settlement and dismissal filing: https://techjusticelaw.org/wp-content/uploads/2026/03/Garcia-v.-Character-Technologies-Inc.-et-al-Entry-115.pdf
• Seventh Circuit vacatur of the Clearview AI settlement (13 July 2026): https://blogs.duanemorris.com/classactiondefense/2026/07/14/seventh-circuit-undoes-novel-privacy-class-settlement-due-to-lack-of-separate-representatives-for-nationwide-class-and-state-sub-classes/
• US Copyright Office, Copyright and AI Part 3 (pre-publication, May 2025): https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-3-Generative-AI-Training-Report-Pre-Publication-Version.pdf
• Thaler v. Perlmutter, No. 25-449 (cert. denied 2 March 2026): https://www.supremecourt.gov/docket/docketfiles/html/public/25-449.html
• GEMA v. OpenAI judgment summary (CMS): https://cms.law/en/deu/legal-updates/gema-vs.-openai-munich-regional-court-i-issues-landmark-copyright-decision