Free tools Windows power users keep installed
One-click scans. No signup required.
Some links on this page are affiliate links: if you buy through them we may earn a commission, at no extra cost to you.
On July 24, 2026, the Delhi High Court refused to stop OpenAI from using ANI news material in the models behind ChatGPT while ANI’s copyright suit proceeds. The single-judge ruling held, on a prima facie basis, that the training use fell within the fair-dealing exception in Section 52(1)(a) of India’s Copyright Act, and that ANI had not shown that ChatGPT memorised or substantially reproduced its reporting. It is an important interim win for OpenAI—not a final ruling that AI training is lawful in India.
The case matters because it separates several questions often blurred together: whether training on copyrighted work is fair dealing, whether a live-retrieval system uses a publisher’s article unlawfully, and whether an AI answer copies protected expression. The court’s answers are provisional and specific to the evidence ANI presented. Read the Delhi High Court judgment.
What ANI alleged—and what it asked the court to do
ANI Media Pvt. Ltd., the news agency Asian News International, sued OpenAI OpCo LLC in the Delhi High Court in CS(COMM) 1028/2024. Its claims raised two distinct copyright questions.
Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Repair Windows errors before they cause bigger problemsFix Now →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →- Training: ANI alleged that OpenAI used its copyrighted news material to train the models behind ChatGPT.
- Outputs: ANI argued that ChatGPT responses could reproduce or closely resemble its protected reporting, and raised concerns about responses attributed to ANI that it said were inaccurate or fabricated.
These are not the same claim. A model’s use of material during development is legally different from a system retrieving an article in response to a prompt, and both differ from an output that reproduces a substantial part of an article. Copyright generally protects original expression, not the underlying facts of a news event. The court found ANI had not established substantial similarity or memorisation at this interim stage; that finding does not establish that no output could ever infringe.
#1 Best Overall
What the court decided
Interim result: Justice Amit Bansal dismissed ANI’s application for an interim injunction on July 24, 2026.
- Jurisdiction: The Delhi High Court held, prima facie, that it could hear the case despite OpenAI’s argument that training took place on servers outside India.
- Training: The court considered OpenAI’s use of ANI material for training to fall within Section 52(1)(a)’s fair-dealing exception, on the record before it.
- Outputs: ANI did not show that the examples were substantially similar to protected expression in its work or proved memorisation and regurgitation.
- Final merits: The suit remains unresolved. The judge expressly said the interim observations would not decide its ultimate outcome.
Why jurisdiction mattered
OpenAI argued that activity on overseas servers put the training conduct beyond the Delhi court’s reach. The court rejected that objection for purposes of the interim application, noting that OpenAI offered its service to users in India and that the allegedly infringing responses had been generated for ANI in India. It found a prima facie basis for jurisdiction under Section 20 of the Code of Civil Procedure and Section 62(2) of the Copyright Act.
That is a meaningful signal for rights holders bringing claims against foreign technology companies: overseas infrastructure alone may not defeat an Indian case where a service is directed to the Indian market and alleged harm occurs here. It is not a rule that every dispute involving an overseas AI company can automatically be heard in India; jurisdiction remains dependent on the facts and the applicable legal tests.
Training is not the same as live retrieval
Training is the process of using material during model development. In simplified terms, the model adjusts its internal parameters based on examples; that is different from storing an article as a document ready to be shown to a user. Whether copies made during training are lawful is the copyright question addressed by the court’s Section 52 analysis.
Rank #2
Retrieval-augmented generation (RAG) works differently. A system can fetch information from an external source, such as a current webpage, in response to a prompt, then use that material as context for a generated answer. A retrieved page need not have been part of the model’s original training data.
The court examined ANI examples that appeared to concern articles published after relevant model-training cut-off dates. It reasoned that those answers could not have come from memorisation of that training data and were more consistent with live retrieval. It also found the cited RAG outputs were not substantially similar to ANI’s original works. That analysis helped explain why ANI’s examples did not prove the output claim at this stage.
RAG is not automatically lawful. Retrieving an article can still raise copyright, licensing, contract, website-terms, access-control, or attribution questions, depending on how the material is obtained and what the system does with it.
The Tool Desk
Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Section 52: fair dealing, not a new AI licence
Section 52(1)(a) of the Copyright Act, 1957, provides an exception for fair dealing with a work (other than a computer programme) for private or personal use, including research; criticism or review; and reporting current events and current affairs. It also addresses electronic storage of a work for those purposes, subject to fair dealing. The court treated the provision as central to whether OpenAI’s training use could be justified.
ANI argued that OpenAI’s commercial business should prevent it from relying on a provision that includes private or personal use, including research. The court did not treat commercial purpose as an automatic disqualification. It viewed the training use as an internal process and noted that the material was not being supplied to users in original or tokenised form. Its analysis also considered whether the use competed with ANI’s exploitation of its content, whether ANI had shown commercial prejudice, and the broader public interest.
India’s statute uses fair dealing, not the broader US doctrine commonly called fair use. The judgment noted that Indian courts have applied different approaches to Section 52(1)(a); it did not establish one definitive test for every AI system or dataset. The practical takeaway is a strong interim argument for AI developers on these facts—not a blanket exemption for commercial model training or web scraping.
Why ANI did not obtain an injunction
An interim injunction requires a court to assess whether the applicant has a prima facie case, whether the balance of convenience favours relief, and whether refusing relief would cause irreparable injury. The court concluded ANI had not made out a prima facie case and that the balance of convenience favoured OpenAI and the public interest.
Do these 3 things before closing this tab:
1Fix the driver behind crashes, sound loss and screen glitches2Clear out junk files and repair common Windows errors3Scan for outdated or missing drivers - takes under a minuteThe judge was concerned that an order could affect ChatGPT users in India, have wider consequences for AI developers, and force developers to seek licences from numerous news agencies and other rights holders. The court also considered the potential effect on AI development, including work by Indian developers, and the risk of giving ANI relief close to a final outcome before trial. Those are considerations in an interim-relief decision, not a settled national policy that innovation always outweighs copyright.
The judgment records that ANI had offered OpenAI a content licence for US$7.5 million. The court treated the figure as relevant to whether the claimed injury could be quantified in money. It is not an award of damages, nor does it establish that ANI’s full claim was valued at that amount.
The court also noted that ANI could block its website from OpenAI’s crawlers, and that OpenAI said it had blocked ANI’s website for training and ChatGPT search/RAG. ANI argued that subscribers or third-party websites could still expose its material. Blocking may reduce some future collection, but it does not by itself undo historical use, remove disputed copies, or settle whether AI answers substitute for licensed news products.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What the ruling means for publishers
The decision gives AI developers breathing room in the short term, but it does not remove publishers’ legal or commercial options. A publisher’s strongest copyright evidence will generally concern identifiable protected expression—not just the fact that a model knows or repeats a public event.
PC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware match- Monitor outputs: Preserve prompts, responses, dates, URLs, and copies of source articles. Compare wording and structure, not only facts, and distinguish a close reproduction from a summary or factual overlap.
- Document economic harm: Evidence of lost subscriptions, traffic, advertising, syndication revenue, or a competing product can be relevant. The court noted that ANI had not placed material before it showing loss of subscribers or syndication business.
- Use technical controls thoughtfully: Crawler restrictions and bot controls can limit some access, but may also reduce visibility in search or AI discovery. They are not guaranteed to prevent all copying or use by third parties.
- Negotiate licences: Licensing may create revenue and clearer terms, though it can favour large companies able to pay and does not itself settle the law for other uses.
- Consider other legal and policy routes: Depending on the facts, contracts, paywalls, access controls, attribution issues, and advocacy for opt-outs or remuneration rules may matter alongside copyright claims.
Public availability is not the same as freedom from all restrictions. Online work can remain copyrighted and may be subject to contractual terms, paywalls, database restrictions, or access controls.
Best Value
What it means for Indian AI startups
For Indian developers, the ruling reduces some immediate uncertainty around training on publicly accessible material, but it does not eliminate litigation or compliance risk. It also highlights why provenance matters: companies that can identify data sources, record model cut-off dates, separate training from retrieval, and test for memorised or verbatim outputs will be better placed to explain their systems and respond to disputes.
Startups still need to assess whether data was obtained under terms restricting automated use, honour applicable crawler restrictions, maintain records, and establish processes for takedown requests, corrections, and misattribution. Licensed or synthetic datasets may be commercially preferable for some high-risk uses even where a fair-dealing argument is available. Legal permission, commercial defensibility, technical ability to remove or filter material, and creators’ acceptance of a system are separate considerations.
The judgment also records a concern that requiring a licence for every work used in training could make development difficult. That argument may shape future policy debate, including proposals for collective licensing or a specific remuneration framework. The interim ruling itself creates no such framework.
What the ruling does not settle
- It is not a final determination of whether OpenAI infringed ANI’s copyright; the suit continues.
- It does not give every AI company permission to scrape or train on all publicly accessible content.
- It does not declare every AI output non-infringing or resolve claims involving verbatim reproduction, inaccurate attribution, or different evidence.
- It does not erase copyright in news reports: facts and original expression are different things.
- It does not definitively resolve uses of paywalled, confidential, or contractually restricted material.
- It does not bind every court in every future dispute or guarantee that an appellate court will reach the same result.
For publishers, the practical case to make may depend on showing what was copied, how it was accessed, and what economic harm followed. For AI developers, this is meaningful interim support for a Section 52 fair-dealing argument, not immunity. OpenAI won the interim battle; the copyright case—and the wider question of how creators should be compensated when AI systems benefit from their work—remains open.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
