AI law and the newsroom, October 16–20: Canadian publishers sue Anthropic, Meta keeps the crawler, and Korea's copyright bills stall
One issue of AI law as it reaches journalism: four Canadian publishers sue Anthropic over crawling and training copies, Meta cuts the word "train" from its 30 October terms while keeping the right to crawl, Australia's AI committee asks the government to look at copyright, and Korea's disclosure bills stall while Google buys Korean books.
This issue covers 16 to 20 October 2026, the gap since the last one, in four items. Four Canadian news companies sued Anthropic in Ontario. Meta removed the word "train" from the terms its business accounts must accept on 30 October, and kept its right to crawl. Australia's AI committee told the government to look at copyright. Korea's training-data bills stayed where they were while Google went shopping for Korean books.
Three questions run through the items below. 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?
Four Canadian publishers sue Anthropic in Ontario
The Canadian Press, The Globe and Mail, Postmedia Network, PNI Maritimes, the Toronto Star and Metroland Media filed a statement of claim against Anthropic and its Canadian subsidiary in the Ontario Superior Court on 6 October 2026. 1
The trade press reported the filing on 14 October. CBC and The Canadian Press carried it on 19 October, which is when most readers saw it. 12
"Anthropic has capitalized on the commercial success of its AI models, building an expansive suite of services, and raised significant capital — all without obtaining a valid licence from any of the News Media Companies," the claim says. It alleges that Anthropic was "substantially and unjustly enriched" at the publishers' expense. 2
The claim pleads four things: copyright infringement, breach of the publishers' website terms, circumvention of the technological measures that restrict copying, and scraping that continues when a user prompts the model. 12
Circumvention is its own head of liability in Canada, separate from copying. The claim names the ClaudeBot crawler and says it went around paywalls, personal-use restrictions, network protections and login authentication. 1
The fourth head reaches past training. The claim says Anthropic "also scrapes content on an ongoing basis in response to user prompts", which would put the retrieval a reader triggers inside the case. 2
The claim counts the works it covers from 2021 onwards, and divides them into what each publisher owns and what it licenses from someone else. 1
| Plaintiff | Works named | Owned | Licensed |
|---|---|---|---|
| The Canadian Press | 4,200,000 | 4,200,000 | — |
| Postmedia and PNI Maritimes | 3,500,000 | 1,750,000 | 1,750,000 |
| Torstar (Toronto Star, Metroland) | 3,500,000 | 1,600,000 | 1,900,000 |
| The Globe and Mail | 2,300,000 | 1,000,000 | 1,300,000 |
That is about 13.5 million works, and about half of them are licensed rather than owned, so a share of any award would pass to other rightsholders. 1
The publishers ask for general damages, an accounting and disgorgement of profits, or statutory damages of up to C$20,000 per work. They also ask for punitive damages and a permanent injunction against further scraping without written consent. 1
Section 38.1 of the Copyright Act sets that ceiling: for commercial infringement, statutory damages run from C$500 to C$20,000 for each work. 3 The claim states no total. No court has ruled on any part of it, and Anthropic had not filed a statement of defence when the filing was reported. 12
The same companies, with CBC/Radio-Canada, sued OpenAI in an Ontario court in 2024, and that case is still unresolved. 2 Canada's fair dealing exception is confined to purposes the statute lists; section 29 names research, private study, education, parody and satire. 4
For a newsroom that is not a plaintiff, the useful part is the pleading. The circumvention and contract claims give publishers routes to a remedy that do not depend on how a Canadian court treats training copies, a question no Canadian court has decided. 1
Meta drops the word "train" and keeps the crawler
This item is a contract amendment, not a regulator's act. Nothing binds Meta or a publisher here except the terms themselves, and no regulator has ruled on them.
Meta published a revision to the commercial terms that take effect on 30 October 2026. Nieman Lab and The Verge reported it the following day, 17 October. 56
The sentence that went read: "Meta may use any content from such domains, web pages, or URLs to provide, develop, improve, research, train, and maintain the integrity of, the Meta Company Products, services, and features provided under these Commercial Terms and any other applicable supplemental terms." 57
The permission to crawl stayed. The text now on Meta's preview page authorises it to use "automated software (such as web scrapers, bots, robots, spiders, crawlers, and user agents)" and to "create test credentials" in order to "access, index, cache, analyze, collect, retrieve, or crawl" the domains and pages "associated with your products or services". 8
What Meta may do with what it takes is now narrower on its face. The crawler may pull updates "in order to provide, develop, improve, research, and maintain advertising and business products" and to "maintain the integrity and security of Meta Company Products". 8
Meta's account is that nothing about its crawling changed. A spokesperson said the consolidated terms "only applied to our ads and business products", that the revision "does not change how Meta uses web crawlers", and that "content owners can use robots.txt to signal their preferences". 6
Three lawyers read the revision for Nieman Lab and were not reassured. Matt Topic of Loevy & Loevy said "develop" is "broad enough to include 'train'", and that "presumably Meta is developing AI business products". James Grimmelmann of Cornell Law School said the new wording "is still broad enough to drive a truck through". 5
The text itself says nothing about honouring robots.txt. Benjamin Sobel of the University of Wisconsin compared the file to a homeowner's sign reading "Don't look at my house": it states a preference and creates no right. 5
The terms take effect on 30 October 2026 and there is no opt-out. A business account that does not agree must stop using Meta's products for business purposes, which for a newsroom means the page and the ad account. 8
Trade bodies answered within days. The News/Media Alliance called the terms "outrageous and probably illegal" on 16 October, and said they "cannot be allowed to be implemented on October 30". 9
The UK News Media Association's chief executive, Theo Bamber, called the terms "the worst kind of arrogant tech platform bullying" and asked ministers and regulators to intervene. 5
On 19 October the campaign group Movement for an Open Web advised publishers to delete their Meta business accounts — Pages, Instagram business accounts, Ads Manager, Business Suite and WhatsApp Business — before 30 October. The group said it is preparing complaints to the European Commission and the UK Competition and Markets Authority. 10
The choice a newsroom faces is a trade, and the terms set its price. It can keep the audience that arrives through Facebook and Instagram and accept the crawl, or leave before 30 October and lose that distribution. The revision narrowed what Meta says it may do with what it takes; it did not restore the exit that was never there.
Australia: the committee tells the government to look at copyright
The Joint Select Committee on Artificial Intelligence wrote to the prime minister this week asking the government to "examine current copyright settings" while making sure creators keep control of their work and are paid for it. Its Labor chair, Jo Briskey, and its Liberal deputy chair, Jonathon Duniam, signed jointly. 11
The committee reports by 30 November 2026. 12 Nothing this week carries draft legislative text, so the report is the next document that could hold a proposal.
The International Confederation of Music Publishers put its case on 19 October. Its managing director, John Phelan, said AI has already been trained on works by Kylie Minogue and Nick Cave without payment, and that of 62 countries that considered changing copyright law for AI, Australia is the one still considering it. 11
The body's evidence was that Google's Gemini reproduced lyrics by Australian artists, and it pointed to a training dataset of more than 12 million songs drawn from YouTube. Google was approached for comment. 11
The committee's members described the position as a stalemate. Models are already trained on copyrighted works in the United States without payment, and companies say they hold back frontier training in Australia while its law is unsettled. 11
Senator David Pocock has said the committee's position on copyright was settled in advance by the prime minister's office, after staff edits appeared in the letter. Ms Briskey rejects that. 11
The AI companies gave evidence too. Anthropic's special envoy, Jeffrey Bleich, told the inquiry the company was not trying to dictate terms, and that "you can't license the entire internet". OpenAI's chief of economic policy, Adam Cohen, pointed to its agreements with News Corp and the Guardian. 11
Nothing binds a newsroom yet. The item is a date to watch and a position to argue: submissions closed on 14 September 2026, and the committee's report is due on 30 November. 12
South Korea: the bills stall while Google buys books
Google has been negotiating with Korean publishers to buy books for AI training since the first half of 2026, and publishing-industry sources told Seoul Economic Daily that some contracts are already complete. Google Korea did not answer a request for comment. 13
Seoul Economic Daily reported it on 18 October 2026. Which publishers signed, and on what terms, could not be confirmed; academic publishers were reportedly among them. 13
The AI Framework Act has been in force since 22 January 2026. Its Article 31 requires a provider of generative AI to tell users in advance that a product runs on it, and to mark what the model produces. 14
Nothing in it requires a provider to say what it trained on. Several amendments would add that, with a duty to disclose the sources of training data and a procedure for rightsholders to check them. They have been introduced and are still pending in the National Assembly. 13
Korean press bodies have asked for the clause since before the Act took effect. The Newspaper Association of Korea sent the science committee and two ministries a proposed amendment adding the duty in October 2025. 15
Until one passes, there is no legal basis in Korea to regulate a distributor that sells content for AI training without the copyright holder's consent. 13
Industry officials warned about price as well. Anthropic's US$3,000-per-work settlement in the Bartz case was damages for piracy, not a fee for training, and treating it as a benchmark risks hardening it into the market rate. 13
One official at a Korean conglomerate developing models said the company does not train on Korean literary works because the data costs more than general data, and that it trains mainly on data supplied by the government. 13
The compliance clock
| Obligation or deadline | Who it binds | Date | Source |
|---|---|---|---|
| Meta Commercial Terms, as revised: crawling and "test credentials" over the pages associated with a business account | Anyone using Meta's products for a business purpose, including every newsroom with a page or an ad account | From 30 October 2026 | Meta Commercial Terms |
| European Commission consultation on the effect of technology on copyright | Anyone responding to it | Closes 3 November 2026 | European Commission |
| Bartz v. Anthropic settlement: first round of payments | Class members whose works have no allocation dispute and whose payment details are complete | On or before 15 November 2026 | Anthropic Copyright Settlement |
| Joint Select Committee on Artificial Intelligence: final report | Parliament of Australia | 30 November 2026 | Parliament of Australia |
| EU AI Act transparency: label deepfakes and AI-generated text on matters of public interest, and mark synthetic media so it can be detected | Providers and deployers of generative AI systems serving the EU | Applying since 2 August 2026 | European Commission |
| EU AI Act: prohibitions on systems that generate non-consensual intimate material and child sexual abuse material, and the marking duty for systems placed on the market before 2 August 2026 | Providers of those systems | From 2 December 2026 | European Commission |
| California: AB 2713 on platform provenance data, SB 1159 on AI systems as non-persons, SB 951 on technology-displacement notices, SB 574 on attorneys and arbitrators | Platforms above two million monthly users; California agencies; covered employers; attorneys and arbitrators | From 1 January 2027 | California Legislative Information, AB 2713 |
The next hard dates
The nearest date is 30 October 2026, when the Meta terms apply and every newsroom with a business account has to have decided. 8
Then 3 November 2026, when the European Commission's consultation on technology and copyright closes 16, and 30 November 2026, when the Australian committee reports. 12
Then 2 December 2026. The AI Act's prohibitions on systems generating non-consensual intimate material and child sexual abuse material apply, and providers of systems already on the market must meet the marking duty. 17
Two files carry no date. Judge Sidney H. Stein has not ruled on the summary-judgment motions in the consolidated OpenAI copyright cases in New York. New York's Stealth Crawler Prohibition Act and FAIR News Act are still unsigned on Governor Kathy Hochul's desk. 18
The Ontario claim is younger and has not started moving either. Anthropic has filed no statement of defence, and no hearing date is set. 1
References
- 1Canadian news publishers sue Anthropic over AI copyright infringement
broadcastdialogue.com
- 2
- 3Copyright Act, section 38.1
laws-lois.justice.gc.ca
- 4Copyright Act, section 29
laws-lois.justice.gc.ca
- 5
- 6
- 7
- 8
- 9News/Media Alliance Statement on Changes to Meta Commercial Terms of Use
newsmediaalliance.org
- 10
- 11
- 12
- 13
- 14
- 15
- 16Commission seeks feedback on challenges and way forward in the area of effect of technology on copyright
digital-strategy.ec.europa.eu
- 17The enforcement framework of the AI Act
digital-strategy.ec.europa.eu
- 18New York State Senate Bill 2025-S9934A
nysenate.gov
This story was produced automatically by a channel. One sentence is all it takes for Neodrop to keep producing for you.