AI Copyright Lawsuits in 2026: What the Big Cases Mean for Businesses Using AI

The short answer

The major AI copyright lawsuits of 2026 are mostly about how AI companies got and used training data, not about ordinary businesses that use AI tools. The clearest results so far: pirating books is expensive (Anthropic's US$1.5 billion settlement), copying a competitor's material to build a rival product is not fair use (the Ross appeal), and whether training generative AI on copyrighted work is lawful is still unresolved. For a business, the practical risk sits in what you publish: outputs that copy protected work, and AI-made content you assume you own. This is general information, not legal advice.

The AI copyright lawsuits that matter, and where they stand

There are dozens of cases. These are the ones that shape the rules, with their status as of October 2026.

Bartz v. Anthropic: a US$1.5 billion settlement

Authors sued Anthropic over books used to train its Claude models. In June 2025 Judge Alsup found that training itself was fair use but that storing pirated copies was not, according to a Norton Rose Fulbright case update. The class action then settled for US$1.5 billion. The court granted final approval in July 2026, as the Association of American Publishers announced; it says Anthropic downloaded nearly half a million books from pirate sites and must destroy those files. Writer Beware estimates roughly US$3,100 per claimed work, with appeals filed in August that could delay payments.

Thomson Reuters v. Ross: the first appeals court ruling

Ross Intelligence used Westlaw headnotes to build a competing legal research tool. On 29 September 2026 the US Court of Appeals for the Third Circuit affirmed that the headnotes are protected and that Ross's copying was not fair use. As Ballard Spahr's analysis notes, the court stressed that "ROSS's AI platform cannot generate original expression", and pointed to harm to Westlaw's market and to a developing market for licensing content as AI training data. It expressly did not decide the generative AI question.

Kadrey v. Meta and the split on market harm

In June 2025 Judge Chhabria found Meta's training on books was fair use on the record before him, while rejecting the idea that training is just like human learning, because a model can produce competing works at scale. Ballard Spahr notes the district judges in Bartz and Kadrey disagree on whether AI training is like human learning, and "no federal appellate court has resolved that disagreement." Remaining claims in Kadrey about distributing ("seeding") pirated files were still being fought as of Norton Rose's update.

The New York Times and other news cases against OpenAI

The news publishers' cases against OpenAI and Microsoft are consolidated in New York. In October 2025 the court declined to dismiss claims about infringing outputs, and in 2026 ordered production of tens of millions of ChatGPT output logs, per Norton Rose. As of October 2026 there is no ruling on the merits.

Outside the US

In the UK, Getty Images v. Stability AI ended largely in Stability's favour in November 2025. The High Court held the model weights were not an "infringing copy" because they do not store the images, and found only limited, historic trade mark infringement tied to watermarks. Getty dropped its training claim because the training did not happen in the UK, as Mishcon de Reya explains.

In Canada, news publishers including the Toronto Star, the Globe and Mail and CBC are suing OpenAI. In Toronto Star Newspapers Limited v. OpenAI Inc., 2025 ONSC 6217, the Ontario Superior Court held it has jurisdiction over six OpenAI entities and refused to send the case to the US; the merits are undecided and the jurisdiction ruling has been appealed, according to Richards Buell Sutton.

What this means if your business uses generative AI

Almost all of these cases target model builders. That does not mean users are risk-free. Your exposure sits in three places.

1. What you publish

If an AI tool produces text, an image or music that closely copies a protected work and you publish it, the infringement is in your marketing, not the model maker's training set. The news cases against OpenAI include claims about outputs, which the court refused to dismiss. Avoid prompts that ask for "in the style of" a living artist or "the lyrics of" a song, and check anything that looks suspiciously polished or familiar.

2. What you think you own

In the US, copyright requires a human author. The Supreme Court declined to hear Thaler v. Perlmutter in March 2026, leaving in place the ruling that a work generated by AI alone cannot be registered, per Norton Rose. Practically, a logo or tagline generated entirely by AI may be hard to protect. Human selection, editing and arrangement strengthen your claim. For brand assets, involve a designer; our branding and logo design work is built for exactly this.

3. What your vendors promise

Check whether your AI vendor's business terms include an indemnity for copyright claims over outputs, and what conditions attach to it, such as using a paid plan or leaving content filters on. Keep records of prompts for important assets, and when an asset really matters, favour tools that say their training data is licensed.

A short risk checklist

• Use business plans of reputable AI tools and read their IP and indemnity terms.
• Do not prompt for named artists' styles, characters, lyrics or articles.
• Run reverse image searches on AI images used in ads or packaging.
• Have a human meaningfully edit anything you need to own: logos, slogans, flagship copy.
• Keep prompts and drafts for key assets as a record of human contribution.
• Treat AI-written blog content like any draft: fact-check and add first-hand value, as we explain in does Google penalise AI content.

Put these rules in writing so the whole team follows them; our AI policy template has a section for it.

Key takeaways

• The headline AI copyright lawsuits target AI developers' training data, not everyday users.
• Pirated training data proved costly; generative AI training itself is still unresolved on appeal.
• Your main risks are publishing infringing outputs and assuming you own purely AI-generated work.
• Human editing, careful prompts and vendor terms are your practical protection.

Frequently asked questions

Can my business be sued for using AI-generated content?

Yes, if the content you publish infringes someone's copyright or trade mark, whoever or whatever created it. The risk is low for ordinary text and original images, and higher when outputs closely resemble a known work, character or brand.

Do I own the copyright in images I make with AI?

In the US, not for material generated by AI alone; copyright requires human authorship. Your own selection, arrangement and editing can be protected. Other countries, including Canada, have not settled this, so get advice for important brand assets.

Has any court said training AI on copyrighted work is illegal?

Not for generative AI at the appellate level as of October 2026. The Third Circuit found Ross's use was not fair use, but stressed that Ross's tool could not generate new expression, and US district courts have split on generative models.

Sources

1. AI in litigation series: An update on AI copyright cases in 2026 — Norton Rose Fulbright, March 2026
2. Third Circuit Addresses Fair Use in AI Training, But Leaves Generative AI Questions Unresolved — Ballard Spahr, 2 October 2026
3. AAP Welcomes Court's Final Settlement Approval in Bartz v. Anthropic — Association of American Publishers, July 2026
4. Anthropic Settlement Update: Final Settlement Approved — Writer Beware, 23 July 2026 (updated 2 September 2026)
5. Getty Images v Stability AI: Unpacking the High Court's judgment — Mishcon de Reya, 6 November 2025
6. Ontario Court Confirms Jurisdiction in OpenAI Copyright Dispute — Richards Buell Sutton, 17 April 2026

How Flowmatix can help

We put this into practice for businesses in Toronto and around the world through our Content & Video Production, ai agents & ai solutions, seo & performance optimization services. Want to know what it would take for your business? Get a personalised quote in about 3 minutes.