AI law and the newsroom, October 2–9: USA Today sues OpenAI, Australia's ABC rejects the opt-out, and a US court pauses Minnesota's AI-image ban

This issue covers 2 to 9 October 2026, the span since the last one, in five items from three countries. The week belonged to one question, asked in three different rooms: on what terms may an AI company take work a publisher owns? In New York, USA Today Co. and the companies that hold its nineteen titles sued OpenAI for more than $250 million. In Sydney, Australia's national broadcaster told a parliamentary inquiry that the opt-out AI companies offer as a compromise cannot work. In Tokyo, a book wholesaler confirmed it had sold books in the thousands to Anthropic for training. The week also produced two rule decisions inside the United States, one from a federal appeals court and one from a state legislature. No new obligation landed on a newsroom this week; four of the five items are about the terms on which AI companies may use what publishers and authors produce.

Three questions run through the items below, and this digest keeps asking them. What may AI companies do with your work? What must you tell your audience when you use AI? Which office, and which document, will ask you about it?

USA Today Co., together with thirteen companies that own its titles, sued OpenAI in the US District Court for the Southern District of New York on 8 October 2026. The complaint is docketed as USA Today Co., Inc. v. OpenAI Foundation, No. 1:26-cv-08892, names seven OpenAI entities as defendants, and says the plaintiffs own the copyrights in material published by nineteen newspapers, among them USA TODAY, The Tennessean, The Arizona Republic, the Detroit Free Press, The Des Moines Register and The Courier-Journal. 1

The plaintiffs put the damage at more than $250 million, and say OpenAI's models have copied hundreds of thousands of their articles and other material. On the statute, they calculate they may recover up to $150,000 for each willful infringement of copyright and up to $25,000 for each act of stripping copyright management information. 1

The complaint raises three claims: copyright infringement under 17 U.S.C. § 501, vicarious infringement, and removal of copyright management information under § 1202(b) of the Digital Millennium Copyright Act. 12 The third claim is the specific one. The plaintiffs say OpenAI built its training data with content extractors it names Dragnet, Newspaper and Gutentag, and that those extractors, "by design," removed the bylines, copyright notices and © symbols from the pages they took. OpenAI knew, the plaintiffs say, that the result would be that those markers reached neither the model nor the reader, which concealed the copying and would make it harder to detect later. 1

The same day, the plaintiffs filed a statement that their action is related to In re OpenAI, Inc. Copyright Infringement Litigation, the consolidated OpenAI copyright cases before Judge Sidney H. Stein in the same court. 3 That is the practical point for a newsroom: the filing does not open a new front so much as add a bloc of nineteen titles to a case already at the summary-judgment stage, where the briefing is complete and the judge has not ruled. The Verge, Forbes and Reuters reported the filing the same day. 24

On 2 October 2026 the US Court of Appeals for the Eighth Circuit granted xAI's request for an injunction pending appeal and paused enforcement of a Minnesota law that bans AI "nudification," the use of AI to turn a photograph of a real person into a sexual image. The court's order ran to a single sentence and gave no reasoning. 56

The Minnesota statute carries civil penalties of up to $500,000 for companies that create non-consensual sexualised images with AI, and lets the people depicted seek damages. The legislature passed it almost unanimously in its 2026 session and Governor Tim Walz signed it in May; it was to take effect on 1 August. 5 xAI sued three days before that date, arguing that the statute "imposes an overbroad, content-based ban on free speech and the tools of visual expression," and asked for a preliminary injunction. A district court first allowed the law to be enforced while the case ran; the Eighth Circuit's order reverses that for now. The case is X.AI LLC v. Ellison, No. 0:26-cv-03425, before Judge Donovan W. Frank in the District of Minnesota, and is now on interlocutory appeal to the Eighth Circuit. 57

Minnesota's attorney general, Keith Ellison, is defending the law, and his office said it disagreed with the ruling and would keep arguing the appeal. 5

This law binds nobody in a newsroom. It reaches the people who run these tools, not the people who publish. The item earns its place for two reasons. It is the first time an appellate court has put a state law regulating what an AI system produces on hold, which shows how far the constitutional objection to AI-content rules has traveled. And the conduct the Minnesota law targets is the same conduct the European Union's AI Act will prohibit from 2 December 2026, when its ban on AI systems that generate non-consensual intimate material applies (see the clock below). Two jurisdictions now regulate the same subject, and one of the two rules is unenforceable while an appeal runs.

The Australian Broadcasting Corporation gave evidence on 6 October 2026 to the Joint Select Committee on Artificial Intelligence, the federal parliamentary inquiry into AI that closed submissions on 14 September. Kate Gilchrist, the ABC's head of content and legal operations, told the committee in Sydney that the broadcaster's journalism had probably already been used to train AI models, and that the compromise AI companies propose, an opt-out encoded in a website's robots.txt file, cannot deliver what it promises. "We cannot scour the internet and ensure that we are opting out on all those sites. It simply does not work," she said. 8910

OpenAI's chief strategy officer, Jason Kwon, gave evidence at the same hearing, and Anthropic appeared as well. 911 Anthropic's written submission accepts that a broad text-and-data-mining exception has been ruled out by the government and proposes instead what it calls a "narrow form of conditional approval" for training, conditioned on rights-holders being able to block crawlers through robots.txt. The ABC and SBS, the other public broadcaster, told the committee that this rests on something that does not hold. "Robots.txt mechanisms are routinely ignored or bypassed," SBS wrote in its submission, and the burden should fall on AI companies to avoid wrongdoing rather than on publishers to monitor every crawler. 11 Research published in the same week found AI agents bypassing robots.txt signals despite their operators' public statements that they respect them. 9

What the broadcasters asked for is not an opt-out but inclusion. Both want AI companies brought inside Australia's News Bargaining Incentive, the levy that since 1 July 2026 has required Meta, Google and TikTok either to strike commercial deals with Australian news publishers or to pay 2.25 per cent of their Australian revenue. That levy was written for search and social platforms; AI companies were placed in a separate category and left out, although they are now among the largest consumers of published journalism. 911

An inquiry takes evidence and passes nothing. The Australian government ruled out a text-and-data-mining exception in July and has said it will legislate protections for creatives, but no bill was before the parliament this week, so no part of this item binds a publisher or a platform. 11

Nippan Group Holdings, one of Japan's largest book wholesalers, has confirmed in writing to the Japan Book Publishers Association that its subsidiary Nippan IPS struck a deal with Anthropic in November 2024 and sold books to the company between December that year and December 2025. The letter, reported on 7 October, said the subsidiary had asked about the purpose of the purchases before agreeing to them but had not known the books were for AI training, and the association shared the letter with some 380 member publishers. 12

The sales came to light through an internal Anthropic document disclosed in a US copyright suit, which named a deal with "Nippan Japan" as part of a company project called Panama that scans books from around the world. Used-book shops in Tokyo took comparable orders between July and September, some for hundreds or thousands of volumes, with shipping addresses in Okayama prefecture. 1213

Japan's copyright law is why the purchases were, on their face, lawful. Article 30-4 of the Copyright Act permits text and data mining for analysis without the rights-holder's consent. Tatsuhiro Ueno, a Waseda University professor of intellectual property, told Nikkei Asia that using physical books for AI training "is legal under Japanese copyright law," and that where the training happens in the United States, US law governs: "as long as the content of the books themselves isn't reproduced, then it's likely to be legal in the U.S. as well, based on recent precedent." 13

Anthropic has already settled one US case about training copies, agreeing in September 2025 to pay US$1.5 billion to book authors who said it had downloaded pirated copies of their work. 9 The Japanese publishers' objection this week is not that a court has ruled against them, because no one has sued. It is that they were not asked. This item binds nobody and changes no rule; it is here because it is the same conflict as the Australian inquiry and the USA Today complaint, run against print rather than newsprint, in a jurisdiction whose law permits the taking.

Pennsylvania's House of Representatives gave final passage on 7 October 2026 to Senate Bill 806, by 163 votes to 40; the bill now goes to Governor Josh Shapiro. 1415

The bill requires advertisers to disclose the use of AI where material generated or modified by AI misleads consumers about the "nature, origin or authenticity of the goods or services being advertised." It also sets formatting duties: an audio disclaimer must be at a volume and cadence a listener can hear, a visual disclaimer must stay on screen long enough to be read, the disclaimer must be given in the same medium as the advertisement and, where feasible, should be permanent or extremely difficult to remove from the synthetic content. The sponsor, Senator Nick Pisciottano, a Democrat from Allegheny County, said the aim was to stop AI being used to create a false impression about a product or a service. 15

The bill is not law. The Senate passed it in June and the House vote was final passage, so it binds anyone only if the governor signs it. The technology trade group NetChoice opposed it in a letter dated 4 October, arguing that federal and state law already prohibit deceptive advertising and that prescriptive formatting rules risk overwhelming short videos, small display ads and audio spots. 15

This is not a rule a newsroom must follow as a publisher, and it was not written for news: it is an advertising-disclosure bill. It belongs in this digest because it is another state-level mandate about what audiences are told when they look at AI-made material, which is the axis California legislated on in September when it required advertising that uses synthetic performers to say so. 15

The nearest dated item is 3 November 2026, when the European Commission's consultation on the effect of technology on copyright closes; the Commission opened it on 29 September. 16 Then 2 December 2026 brings the EU's prohibitions on AI systems that generate non-consensual intimate material and child sexual abuse material; that is also the date by which providers of AI systems placed on the market before 2 August 2026 must comply with the AI Act's marking and detection duty for generated content. 17 1 January 2027 carries four California obligations — AB 2713, SB 1159, SB 951 and SB 574 — and Connecticut's chatbot-disclosure duty; New Mexico's legislature opens its session the same month, with the Frontier Artificial Intelligence Safety and Accountability Act to be filed.

Four files in this issue still have no date. Judge Stein has not ruled on the summary-judgment motions in the OpenAI MDL, and the USA Today action will take its place in that structure. Australia's committee has not set a reporting date. New York's two AI-and-news bills are still unsigned: the Stealth Crawler Prohibition Act, passed in June, sits on Governor Kathy Hochul's desk, and the FAIR News Act, which would require a label on news content substantially created by generative AI, passed both chambers and is waiting on the same desk. 1819 And Japan's publishers' association is waiting for Nippan to answer the questions its members have put about the sales.