Home โ€บ Blog โ€บ Data law and the data wall
Inside the Data Layer ยท Part 8 ยท Law and limits

Who Owns the Training Data? Copyright, Consent and the Coming Data Wall

Two forces now shape the data layer from outside: the law and scarcity. Courts in the United States, Britain and Germany have started ruling on whether training on copyrighted work is allowed; Europe now requires model makers to publish what they trained on; and the supply of fresh, high-quality human text is running short. This final part of the series maps the rulings and rules as of October 2026, explains the opt-out signals sites use, and models when the "data wall" arrives and what labs are doing about it.

FL
FrontierAGI Team
โš ๏ธ Not legal advice (as of 27 Oct 2026) This is a journalistic summary from court reporting, law-firm analyses and official texts; laws differ by country and cases are still on appeal. Details not confirmed against a primary source are marked โš ๏ธ. The scenario explorer and data-wall model are editorial illustrations. Disclosure: Claude, made by Anthropic, helps write this site; one of the leading cases discussed here, Bartz v. Anthropic, involves Anthropic, and we report it from public coverage like the others.

1. Case file

๐Ÿ“œ
Data law and limits, October 2026The outer boundary of the data supply chain
Bartz v. Anthropic (US, Jun 2025)Training on bought books fair use; pirated copies not โ†’ $1.5B settlement
Kadrey v. Meta (US, Jun 2025)Meta won on its record, but judge warned training is often not fair use
Thomson Reuters v. Ross (US, Feb 2025)Fair use rejected for a competing, non-generative tool
Getty v. Stability AI (UK, Nov 2025)Model "not an infringing copy"; narrow trademark win for Getty
EU AI Act (from 2 Aug 2025)Public training-data summary + copyright policy for GPAI models
Public human text (Epoch AI)~300T tokens; fully used 2026โ€“2032 at trend

2. Timeline

Law and limits, 2018 โ†’ 2026

Tap a milestone. Colour: legislation, US cases, other courts, scarcity.

Tap a milestone.

3. United States: fair use, case by case

US copyright law has no specific rule for AI training; it turns on fair use, a four-factor test weighing the purpose of the use, the nature of the work, how much was taken, and the effect on the market for the original. Three 2025 decisions set the early tone:

Bartz v. AnthropicJudge Alsup held in June 2025 that training on lawfully acquired books was "exceedingly transformative" fair use, but that downloading millions of pirated books to build a library was not. Anthropic settled the class claims for $1.5 billion (about $3,000 per book for ~500,000 books); the settlement later received final approval, with some authors opting out.
Kadrey v. MetaDays later Judge Chhabria ruled for Meta on summary judgment, but only because the authors had not shown market harm. He wrote that in most cases training on copyrighted works without permission is likely not fair use, stressing that flooding the market with competing works can harm creators even if the use is transformative. Other claims continued.
Thomson Reuters v. RossIn February 2025 Judge Bibas rejected fair use for Ross Intelligence's use of Westlaw headnotes to build a competing legal search tool. The ruling is narrow (the tool was not generative and competed directly) and went to the Third Circuit, which heard argument in 2026 โš ๏ธ.

Dozens of other cases remain open, including The New York Times v. OpenAI and Microsoft and suits by music publishers, artists and newspapers. The pattern so far: how data was acquired matters as much as how it was used, and harm to the market for the original is the battleground.

4. Britain, Germany, Japan

๐Ÿ‡ฌ๐Ÿ‡ง Getty Images v. Stability AIOn 4 November 2025 the High Court in London largely rejected Getty's case. Getty had dropped its training claims during the trial (the training happened outside the UK); the judge held that a model like Stable Diffusion, which does not store or reproduce the works, is not an "infringing copy". Getty won only narrow trademark points over generated watermarks. The UK government consulted in 2024โ€“25 on a text-and-data-mining exception with an opt-out โš ๏ธ.
๐Ÿ‡ฉ๐Ÿ‡ช GEMA v. OpenAIIn November 2025 a Munich court found that ChatGPT infringed song lyrics that the model had memorised and could reproduce, rejecting the argument that EU text-and-data-mining exceptions covered this โš ๏ธ. The decision treats memorised content inside the model differently from the UK's Getty ruling.
๐Ÿ‡ฏ๐Ÿ‡ต JapanArticle 30-4 of Japan's copyright law (2018) broadly allows use of works for information analysis, including AI training, unless it unreasonably harms the rights holder; it is among the most permissive regimes โš ๏ธ.

5. The EU AI Act: disclose what you trained on

Since 2 August 2025 providers of general-purpose AI models in the EU must keep technical documentation, adopt a copyright policy that respects rights-holders' machine-readable opt-outs under the EU's text-and-data-mining rules, and publish a summary of training content using a template the European Commission released on 24 July 2025. The template balances transparency with trade secrets; a voluntary Code of Practice offers a route to compliance.

The training-data summary template

Tap a section. Our summary of the Commission's template.

6. Rulings at a glance

Simplified; many cases have further claims, appeals or settlements. โš ๏ธ = reported, not checked against the judgment.

7. Is it allowed? A scenario explorer

Pick an activity and a place

Our editorial reading of the direction of rulings and rules as of October 2026. Not legal advice; outcomes depend on facts.

8. Opt-out signals

SignalHow it worksLegal weight
robots.txtDisallow specific AI crawlers (GPTBot, ClaudeBot, CCBot, Google-Extendedโ€ฆ) โ€” see Part 2Voluntary in the US; can count as a machine-readable reservation in the EU
Terms of serviceSite terms forbidding scraping or AI trainingContract claims; enforceability varies
TDM reservationEU Article 4(3) opt-out, e.g. in metadata or protocols such as the TDM Reservation Protocol โš ๏ธBinding in the EU for commercial text and data mining
Network blockingCloudflare's default AI-crawler block and pay-per-crawl (Jul 2025)Technical, not legal: crawlers simply cannot fetch
Dataset opt-outs"Am I in The Stack?", artist opt-out registriesVoluntary, per dataset

9. The data wall

Epoch AI estimated in 2024 that the effective stock of public human text, adjusted for quality and repetition, is about 300 trillion tokens (90% range 100Tโ€“1,000T), and that if training datasets kept growing at their historical pace, frontier models would use all of it between 2026 and 2032. Restrictions (Part 2) shrink the usable share further; filtering (Part 3) shrinks it again.

When do we hit it?

Our model: the largest training dataset starts at ~15T tokens in 2024 and grows by a chosen factor per year; compare with the stock estimate and the share you assume is usable. Change the inputs.

10. Ways over the wall

Tap a strategy.

11. The series, complete

Inside the Data Layer: eight parts

VerdictThe data layer has reached its outer walls. Legally, early rulings suggest training itself can be lawful while piracy, memorised outputs and market substitution are not, and Europe has made disclosure mandatory; the appeals that will settle these questions are still to come. Physically, the stock of fresh public human text is close to exhausted for the largest models. The response runs through every part of this series: licensing instead of scraping, careful filtering and repetition, synthetic data anchored to verifiers, expert human data, and the infrastructure to prove where every token came from. Data has become the slowest, costliest and most contested input of the AI factory.

12. Reading the numbers

Primary / widely confirmed
Bartz v. Anthropic: fair use for lawfully acquired books, not pirated copies (Jun 2025); $1.5B settlement, ~$3,000 ร— ~500K books, final approval. Kadrey v. Meta: summary judgment for Meta, judge's warning on market harm (Jun 2025). Thomson Reuters v. Ross: fair use rejected (Feb 2025). Getty v. Stability AI: High Court, 4 Nov 2025, model not an infringing copy, narrow trademark findings. EU AI Act GPAI obligations from 2 Aug 2025; template published 24 Jul 2025. Epoch AI: ~300T tokens (100Tโ€“1,000T), 2026โ€“2032.
Reported โš ๏ธ
Ross appeal status; GEMA v. OpenAI (Munich, Nov 2025); Japan Article 30-4 scope; UK consultation; TDM Reservation Protocol details.
Our illustrations
Scenario explorer verdicts, data-wall model (15T start, chosen growth, usable share and epochs), strategy notes.