10 AI Regulatory and Legal Battles to Watch in 2026

Most AI news is about what the models can do. A growing share of the consequential news is about what they are allowed to do, who pays for the data behind them, and who answers when something goes wrong. This article follows ten of those fights as they stood on October 18, 2026. For each one it gives the court or body, the date, the status, and a plain split between what has been decided and what is only claimed.

What are the ten fights at a glance?

Where do the copyright fights stand?

1. Thomson Reuters v. ROSS Intelligence: did an appeals court say AI training is not fair use?

Parties and forum. Thomson Reuters, owner of the Westlaw legal database, sued ROSS Intelligence, a legal-research startup, in the federal court in Delaware. The Third Circuit Court of Appeals heard the appeal. What was decided. According to IP Watchdog's report of September 30, 2026, the appeals court affirmed the lower court's ruling that ROSS infringed more than 2,000 copyrighted Westlaw headnotes and affirmed the denial of summary judgment to ROSS on its fair-use defense. The report calls it the first major US circuit ruling on fair use in AI training.

Why it is limited. The opinion was issued under seal, with the parties told to propose redactions, so we have not read the court's reasoning. The same report describes ROSS's tool as non-generative, and a commentator quoted in it criticised the reasoning. Whether the holding reaches chatbots that generate new text is not settled by this case. Alleged vs decided: the copyright and fair-use points are decided on that record; any wider effect on other cases is unknown.

2. Bartz v. Anthropic: what does a $1.5 billion copyright settlement settle?

Parties and forum. A class of authors sued Anthropic in the US District Court for the Northern District of California. What was decided. The Authors Guild reported that Judge Araceli Martínez-Olguín granted final approval and entered final judgment on July 20, 2026, for a $1.5 billion settlement.

What it covers. Per the Guild, class members release claims about Anthropic's past acquisition and copying of their works. They do not release claims about the system's outputs, claims about future conduct, or claims over works not on the settlement's list. That means the settlement resolves one set of claims against one company and does not rule that AI training is lawful or unlawful. Alleged vs decided: the settlement and its approval are decided; no court ruled on the underlying fair-use question through this settlement.

3. In re OpenAI Copyright Infringement Litigation: what is happening with the New York Times case?

Parties and forum. The news plaintiffs (including The New York Times) and the class plaintiffs (including the Authors Guild and book authors) are suing OpenAI and Microsoft in cases consolidated before Judge Sidney Stein in federal court in New York, according to a case-tracking blog. Status. The same tracker reported on September 4, 2026 that both sides would file their summary-judgment motions that day. We did not find a ruling, a hearing date or a trial date.

Alleged vs decided: the plaintiffs allege their works were used without permission to train and run the models, and the defendants dispute liability; no ruling on fair use in these cases has been reported. We did not open the docket, so check the court's record before relying on any date. Earlier reports of amended complaints and discovery orders came from low-quality trackers, so we omit them.

4. Getty Images v. Stability AI: why did the UK court mostly side with the AI company?

Parties and forum. Getty Images sued Stability AI in the High Court of England and Wales. What was decided. In its report dated December 19, 2025, the law firm Burges Salmon says the High Court judgment, handed down in November 2025 by Joanna Smith J, was mixed and largely favoured Stability AI. An early version of Stable Diffusion technically infringed Getty's trade marks, but the infringements were historic and limited. Most of the trade-mark claim failed, and so did the claim of secondary copyright infringement. Getty had withdrawn its primary copyright claim because the model was trained outside the UK.

What is next. The firm reports that permission to appeal was granted on the secondary-infringement question. We found no hearing date. Alleged vs decided: the first-instance result is decided; the appeal is pending as far as we could confirm.

Which corporate and consumer fights matter?

5. Musk v. OpenAI: what did the jury decide?

Parties and forum. Elon Musk sued OpenAI, its chief executive Sam Altman and president Greg Brockman in a federal court in Oakland, California, over the company's shift from a nonprofit mission. What was decided. Reuters, as carried by BusinessDay, reported on May 18, 2026 that a jury unanimously found OpenAI not liable after deliberating for under two hours, concluding that Musk brought his case too late. Judge Yvonne Gonzalez Rogers said there was a substantial amount of evidence supporting the finding, and described the timing question as one of fact.

What it does not decide. The reported basis is timing, so the verdict does not say whether OpenAI's conduct matched its original mission. Musk's lawyer reserved the right to appeal. Alleged vs decided: the verdict is decided; the merits were not reached.

6. States v. OpenAI: what are attorneys general alleging?

Parties and forum. The Associated Press reported on October 15, 2026 that at least eight states are suing OpenAI in separate state-court actions, alleging that ChatGPT violates consumer protection laws by harming children and deceiving adults. The report names Arizona, Arkansas, Idaho, Kansas, Montana, Nebraska and West Virginia as filing on one day, and describes an Iowa suit as well.

Alleged vs decided. All of it is allegation; the report mentions no court ruling. OpenAI says it has built child-safety protections into ChatGPT and wants to work with states on policy that applies across the industry, and it makes claims about teen usage that the report does not verify. We cover the same suits, and other documented incidents, in AI safety incidents and warning signs of 2026.

Who gets to regulate AI?

7. Executive Order 14365: can the federal government challenge state AI laws?

What it does. The order, titled Ensuring a National Policy Framework for Artificial Intelligence, was signed on December 11, 2025 and published in the Federal Register on December 16. Its Section 3 requires the Attorney General to set up an AI Litigation Task Force within 30 days, whose sole responsibility is to challenge state AI laws that the order says unconstitutionally regulate interstate commerce or are preempted by federal regulation. Section 5 ties some remaining broadband funding to states not having onerous AI laws, to the extent federal law allows.

What it does not do. An executive order directs agencies; it does not itself overrule state statutes. The order's child-safety carve-out, as we read the text, applies to the legislative recommendation it requests, not clearly to the task force's challenges. Some secondary sources describe task-force lawsuits, but we could not confirm any from a primary source, so we report none. Alleged vs decided: the order is in force; its legal effect on state laws will depend on courts.

8. Colorado SB26-189: what replaced the state's first AI Act?

What happened. The Colorado legislature's bill page shows SB26-189, Automated Decision-Making Technology, signed by the governor on May 14, 2026 as Session Law Chapter 131. It repeals and re-enacts the consumer-protection provisions of the 2024 law, SB24-205, which had set duties for developers and deployers of AI systems. Developer obligations begin on January 1, 2027, and the attorney general must adopt certain disclosure rules by the same date.

What is unsettled. Secondary sources describe a federal court stay of enforcement in a suit brought by xAI, and differ on whether it covers the new law. We could not confirm that from a primary source, so treat enforcement status as unverified and check the court docket and the attorney general's guidance. Alleged vs decided: the new statute is enacted; its enforcement and any court challenge are not confirmed by us.

9. The EU AI Act and AI Omnibus: what changed on dates?

What the Commission says. Its AI Act page gives entry into force as August 1, 2024 and general-purpose AI obligations as applicable from August 2, 2025. The AI Omnibus, adopted November 19, 2025, reached political agreement on May 7, 2026 and entered into force on July 27, 2026. It set Annex III high-risk obligations (such as biometrics, education and employment) for December 2, 2027 and obligations for AI embedded in regulated products for August 2, 2028, and added a prohibition on AI systems that generate non-consensual sexually explicit content, effective December 2026.

Where to go deeper. We explain what the delay means for compliance planning in our guide to the EU AI Act's high-risk dates and the Digital Omnibus. Alleged vs decided: these dates are law as the Commission publishes them; check its page for later changes.

How are AI chips tied up in law?

10. AI chip exports to China: what did the US change in January 2026?

What changed. In a press release dated January 13, 2026, the Commerce Department's Bureau of Industry and Security said it would review licence applications for chips such as Nvidia's H200 and AMD's MI325X on a case-by-case basis, provided security requirements are met. The listed conditions include that exports must not reduce global chip production capacity available to US customers, that the buyer has compliance measures including customer screening, and that the product undergoes independent testing in the United States.

What we cannot say. We did not confirm which licences have been granted or whether any chips have shipped, because the reports we found were secondary or undated. The rule is policy, not a promise of exports. For how chipmakers fit into the wider AI stack, see our overview of AI hardware and chip companies. Alleged vs decided: the licensing policy change is decided; actual exports are unconfirmed.

How can I follow a legal fight without being misled?

For how we weigh claims of different strength, see our methodology and the guide to vendor claims versus verified findings. If we got a status wrong or a case has moved, tell us and we will record it in our corrections.

What are the limitations of this article?

• We are not lawyers and this is not legal advice.
• We did not read court dockets or the sealed Third Circuit opinion, so several case details rest on reports and trackers, which we name.
• Statuses change quickly. Everything here is as of the source dates shown, and we ran no independent checks beyond reading those pages.
• We left out matters we could not confirm from a readable source, including the reported task-force lawsuits, trial timing in the OpenAI copyright cases, and details of chip licences and shipments. Absence from this list is not a judgement that a matter is unimportant.