The Week AI Rules Moved the Decision
On Wednesday, Meta took its new advertising and business terms off the calendar altogether. Three days later, Britain's competition regulator opened a case about the same terms.
0:00 / 5:29
The Week AI Rules Moved the Decision
Meta took its new commercial terms off the calendar on Wednesday, and collected a competition case for them on Friday. A judge in Manhattan now holds every friend-of-the-court brief he is going to get before he decides whether training on news articles is fair use. And Google and Microsoft have both told publishers where to record an AI preference, in a file the newsroom already owns. None of the three takes effect this morning. Here is what each one actually does.
Meta's terms are off the calendar, and the case is on them
Meta announced on Wednesday 28 October that it was postponing its updated commercial terms and self-serve ad terms indefinitely. Its statement said the company had heard publishers' feedback, did not plan to make policy changes without a full consultation with industry groups and leaders, and expected that consultation could extend beyond 20 November. Until then, the existing terms remain in place. 1
That removes the 20 November effective date, and with it the arbitration opt-out window that ran 30 days from first acceptance — the deadline this programme flagged last week at around 20 December. What a publisher faces now is the contract Meta already has.
Two days later the Competition and Markets Authority opened a formal investigation under Chapter II of the Competition Act 1998, the provision that covers abuse of a dominant position. The regulator says it used that power rather than the newer digital markets regime because Meta has not been designated with strategic market status; the next candidates for designation go to the CMA board in spring 2027. 23
The businesses under investigation are Meta Platforms, Inc., Facebook UK Limited and Facebook Ireland Limited. The CMA will assess whether Meta holds a dominant position in one or more UK markets and, if so, whether requiring businesses to accept these terms amounts to unfair trading conditions. The complaints behind the case came from the News Media Association and from the campaign group Movement for an Open Web. The case timetable runs from October 2026 to January 2027 for rapid information gathering, and the CMA can impose interim measures while it investigates. 3
"Where a company holds a dominant position, it must not use that position to impose unfair terms." — Sarah Cardell, chief executive of the CMA 2
The detail that matters for planning: the investigation covers Meta's existing terms as well as any revision. Withdrawing the paused text would not end the case. No decision has been made on whether Meta has broken the law, and the CMA has said no assumption should be drawn at this stage. 2
The briefs that close the OpenAI fair-use record
Friday 30 October was the deadline for friend-of-the-court briefs in In re OpenAI, Inc., Copyright Infringement Litigation, the consolidated case before Judge Sidney Stein in the Southern District of New York. Cross-motions for summary judgment were filed on 4 September, and replies on them are due on 20 November under the scheduling order the judge approved on 24 September. No ruling has been issued. More than a dozen news publishers are among the plaintiffs. 4
The News/Media Alliance filed with the National Newspaper Association, the National Press Photographers Association and ProPublica. Its brief argues that an answer built from a news article substitutes for that article, and that no fair-use precedent covers copying a work in order to compete with it. 4
"This case is existential for the news media industry." — Danielle Coffey, president and chief executive of the News/Media Alliance 4
The Association of American Publishers, STM, the Independent Book Publishers Association and the Copyright Clearance Center filed the day before, supporting the plaintiffs' motion. Their brief cites an AI training licensing market valued at USD 5.2 billion in 2025 and projected to reach USD 55.8 billion by 2035, and argues that "the fair use doctrine is intended to adapt—not yield—to changes in technology." 5
On the same deadline, three Democrats on the House Judiciary Committee — Jamie Raskin, Jerrold Nadler and Hank Johnson — filed against the Justice Department's position. The department had intervened on 1 September to argue that training is transformative and that national security weighs against restricting it. The lawmakers' brief answers that national security "is not a fair use factor under the Copyright Act," and that treating it as one would add an exception Congress did not enact; they describe the department's argument as "a bulldozer" for AI training. 67
The Wikimedia Foundation also filed on 30 October, and split the question differently from everyone else. It argues that training a model can be fair use where the output is something new, but that retrieving a work from the live web at the moment a chatbot answers, and serving it back to the reader, is not fair use and should require a licence — in Wikipedia's case CC BY-SA 4.0. 8
That is the split worth watching. For a newsroom, the cost arrives less often as a training set than as an answer served in place of the article.
Where to record the preference, and what each switch covers
Google said on Friday 30 October that it will support a shared AI training control in
robots.txt, so that one setting expresses a preference to every participating provider instead of one crawler name per company. Sites that opt out through the shared token without naming Google-Extended will be treated as having disallowed Google-Extended. Sites that specifically allow Google-Extended remain eligible to appear in responses and receive clicks through grounding in Gemini Apps. Google aims to support the control in 2027. 9The mechanism comes from an Internet-Draft,
draft-kelley-aipref-training-preference-signal-00, written by Kevin Kelley of Anthropic and filed the same day. The token is __usage_ai-training. The IETF states that the document is not endorsed by the organisation, has no formal standing in its standards process, and expires on 3 May 2027. 10Microsoft went further and split the two uses apart. AI Training is set in
robots.txt or, overriding it, on an individual page. AI Grounding is set page by page only, and has three values: allow, disallow, and DisallowExtended, which permits use inside Microsoft-operated experiences such as AI Mode, AI Overview, Copilot and Microsoft 365 while excluding the grounding services Microsoft provides to third parties, including Web IQ and the grounding API in Foundry. 11| Setting | Where it is written | What it controls |
|---|---|---|
| AI Training | robots.txt, or a page-level tag or header that overrides it | Whether content may be used to train generative models |
| AI Grounding | Page-level only | Whether content may inform generated answers |
DisallowExtended | Page-level only | Microsoft's own products only; not third-party grounding |
The Bing timetable is explicit. The new preferences take effect on 1 February 2027 as Microsoft recrawls. Sites using
NoArchive or NoCache have until 1 March 2027 to publish the new settings, or those tags stop being read as a preference for training or grounding. Temporary grounding transition preferences open in Bing Webmaster Tools on 15 December 2026 for sites using the old tags, and expire on 30 September 2027. Traditional Bing search rankings are unaffected. 11Both Microsoft preferences, and the shared robots.txt token, default to allow. A site that publishes nothing is treated as consenting. Google's announcement also argues that the proposal is consistent with the EU Code of Practice for General-Purpose AI under the AI Act, which recognises
robots.txt as a primary method for publishers to express machine-readable preferences — which is how a voluntary technical file starts to look like the compliance record. 9Dates to hold
- 3 November 2026 — deadline for Ross Intelligence's petition for rehearing en banc in Thomson Reuters v. ROSS, the Third Circuit case that rejected fair use for training on Westlaw headnotes.
- 9 November 2026 — comments close on the AI Office's draft implementing regulation for AI Act Articles 75c and 75d, covering access to file, procedure and fines.
- 20 November 2026 — replies due on the cross-motions for summary judgment in the OpenAI copyright case. It was also Meta's last stated date for its commercial terms, which it now says it may not meet.
- 15 December 2026 — temporary grounding transition preferences open in Bing Webmaster Tools for sites on
NoArchiveorNoCache. - 1 February 2027 — Microsoft begins honouring the new AI Training and AI Grounding preferences as it recrawls.
- 1 March 2027 — deadline for sites on the old tags to publish the new preferences; after this date
NoArchiveandNoCacheare no longer read for training or grounding. - 30 September 2027 — Bing's temporary grounding preferences expire.
- No date — Meta's revised commercial terms.
References
- 1
- 2
- 3
- 4
- 5
- 6
- 7US House Judiciary Committee Democrats, amicus announcement
democrats-judiciary.house.gov
- 8Wikimedia Foundation, files to protect open knowledge in New York Times v OpenAI case
wikimediafoundation.org
- 9Google Search Central Blog, supporting a shared AI training control for websites
developers.google.com
- 10IETF datatracker, draft-kelley-aipref-training-preference-signal-00
datatracker.ietf.org
- 11
This story was produced automatically by a channel. One sentence is all it takes for Neodrop to keep producing for you.