What is coming in India
IndiaEU
- November 2025Rulings on what happens after collection
- December 2025The DPIIT working paper
- 24 July 2026The interim ruling in ANI Media v OpenAI, and its appeal
- 2 August 2026Enforcement of the AI Act
- 3 September 2026The Federal Court of Justice in Kneschke v LAION
- 3 November 2026The Court of Justice in Like Company v Google
- 10 November 2026
- 8 December 2026The interim ruling in ANI Media v OpenAI, and its appeal
- 17 December 2026The Federal Court of Justice in Kneschke v LAION
- May 2027The DPDP Act
- 2 August 2027Enforcement of the AI Act
Past Still to come, as of 10 November 2026
The interim ruling in ANI Media v OpenAI, and its appeal. A Division Bench of the Delhi High Court is hearing ANI's appeal against the interim ruling of 24 July 2026. The appeal is listed for 8 December 2026. If the Bench leaves the ruling in place, the facts it rested on are likely to become the working reference. A team will then need to show for each source that the content was freely available, that no barrier was passed and that the site had not opted out. If the Bench takes a different view, a team will need to know exactly which sources it collected and what each one said. A receipt answers both needs.
The trial of the suit. The interim ruling was made on a first view. The final decision will be made on evidence, and the court has said that its interim observations do not bind it. A position that depends on the interim ruling alone can change.
A first ruling on Section 43 and crawling. No Indian court has yet decided whether crawling a public page against a site's stated wishes is access without permission. When a court does decide, the question will be what the owner permitted and what the collector knew. A sealed record that you looked for the owner's wishes and followed them is the evidence that question calls for.
The DPIIT working paper. The committee's first paper, published in December 2025, proposes a mandatory licence for training on "lawfully accessed" works, with payment to rightsholders and a training data disclosure by developers. A second paper on AI outputs is expected. Any change would need Parliament to amend the Copyright Act. If a model of this kind becomes law, developers will need records of what they trained on and how they reached it.
The DPDP Act. The main duties under the Digital Personal Data Protection Act are due to apply from May 2027. Teams that collect web data containing personal data will need a basis for it that is separate from anything RightSignal decides.
What is coming in the EU
IndiaEU
- November 2025Rulings on what happens after collection
- December 2025The DPIIT working paper
- 24 July 2026The interim ruling in ANI Media v OpenAI, and its appeal
- 2 August 2026Enforcement of the AI Act
- 3 September 2026The Federal Court of Justice in Kneschke v LAION
- 3 November 2026The Court of Justice in Like Company v Google
- 10 November 2026
- 8 December 2026The interim ruling in ANI Media v OpenAI, and its appeal
- 17 December 2026The Federal Court of Justice in Kneschke v LAION
- May 2027The DPDP Act
- 2 August 2027Enforcement of the AI Act
Past Still to come, as of 10 November 2026
The Federal Court of Justice in Kneschke v LAION. Germany's highest civil court heard the case on 3 September 2026 and will announce its decision on 17 December 2026. It has indicated that it may refer questions to the Court of Justice of the European Union. The central question for collectors is whether a reservation written in ordinary language can count. If it can, a team that read only robots.txt has a gap across everything it collected. If it cannot, a team that followed such reservations gave up some sources and lost nothing else.
The Court of Justice in Like Company v Google. The Advocate General gave an opinion on 3 November 2026 in the first case on AI chatbots and press publishers' rights. The Court's judgment will follow, and it is likely to shape how retrieval products may use news content.
The Commission's list of opt-out protocols. The Commission consulted in late 2025 and early 2026 on which machine-readable protocols for reserving rights should be generally recognised. It has not yet published the list. When it does, the list is likely to become the reference for which protocols a provider is expected to read. RightSignal already reads the main protocols in use and adds to them through new signed versions.
Enforcement of the AI Act. The Commission has been able to enforce the duties on general-purpose model providers since 2 August 2026. Public summaries of training content are now being published and read by rightsholders. Models placed on the market before August 2025 must meet the duties by 2 August 2027.
Rulings on what happens after collection. Courts are also examining model outputs. In GEMA v OpenAI the Munich Regional Court held in November 2025 that song lyrics memorised by a model had been reproduced, and that ruling is under appeal. Disputes of this kind begin with the question of what was collected and from where.
Why a strict reading serves you
RightSignal reads unsettled law cautiously, and it does so on purpose.
| Open question | How RightSignal reads it today | If a court takes the strict view | If a court takes the looser view |
|---|---|---|---|
| Does a robots.txt rule against your crawler withdraw permission under Section 43? | It follows the rule and stops. | Your past collection already respected it. | You gave up some sources and nothing else. |
| Does a refusal in a site's terms withdraw permission in India? | It treats a clear refusal as a stop. | Your past collection already respected it. | A new version of the Algorithm can open those sources. |
| Does a reservation written in ordinary language count in the EU? | It follows the reservation. | Your records show that you read and followed it. | A new version of the Algorithm can open those sources. |
| May a product use news content to answer questions? | It holds the purpose for review on news content. | You did not build on sources you must now remove. | You can add the sources once the position is clear. |
The costs on the two sides are not equal. A strict reading that turns out to be stricter than the law costs you some sources for a time. A loose reading that a court later rejects can cost you a dataset, the model trained on it and the confidence of your customers. When the law settles, RightSignal issues a new signed version. Decisions already made are not made again, and their receipts do not change, so your record of what you did under the earlier version stays intact.
Why an Indian team should read EU law now
The AI Act applies its copyright duties to any provider that places a general-purpose model on the EU market, wherever the model was trained. An Indian company that trains under Indian law alone and later sells into Europe will be asked about reservations it never looked for, on sites that may since have changed. The India and EU Bridge profile avoids that position. It applies the stricter of the two readings at every step, so the evidence EU buyers and regulators ask for is gathered while you build for India. You then approach the EU market with that evidence in hand and do not have to rebuild it. The Bridge does not by itself meet any duty under the Act.
Why now is the time to integrate
- Evidence cannot be made afterwards. A receipt records what a site said at the moment you collected. Sites change their terms, and public archives keep little of them. Every month of collection without receipts is a month you will be able to evidence only weakly, whatever the courts decide.
- The questions are already being asked. Model providers serving the EU are publishing summaries of their training content today. Enterprise buyers and investors ask about provenance today. Indian sites cite Section 43 in their terms today.
- Checking first is the cheap order of events. Leaving a source out before a crawl costs nothing. Removing it after a model is trained can mean retraining.
- Each coming decision raises the value of a record. Whichever way the appeal, the German decision and the European judgment go, each of them is likely to make the question of what a site said, and when, matter more.