AI and Copyright: Who Owns AI-Generated Work, and the Cases Being Fought Now

By the Cehpoint teamPublished 6 min read

Short answer: In ANI v OpenAI, the Delhi High Court refused an interim injunction in July 2026, treating training as prima facie fair dealing and finding no proven copying in outputs, though the case goes to trial. US courts require a human author, and India's Copyright Act names the person who causes a computer-generated work to be created as its author. Businesses should keep real human creative input, keep records and check outputs for copying.

Two copyright questions now sit on every marketing team's desk. First, if we publish text, images or code made with AI, do we own it? Second, could the AI tool we used have been built on someone else's copyrighted work, and does that expose us? Courts in India, the US and the UK have started answering parts of both questions, though none has answered them completely.

This article explains the main cases being fought, what they actually decided, and practical rules for a business publishing AI-assisted content today. It is a business guide, not legal advice; for a specific dispute, speak to an IP lawyer.

India's headline case: ANI v OpenAI in the Delhi High Court

The news agency ANI sued OpenAI in the Delhi High Court, alleging that OpenAI scraped, stored and used ANI's news articles without permission to train the models behind ChatGPT, and that ChatGPT's outputs reproduced ANI's work. ANI asked for an interim injunction to stop OpenAI from storing or using its content.

On 24 July 2026, Justice Amit Bansal refused the interim injunction, as reported by LiveLaw. According to a detailed analysis on the Cyril Amarchand Mangaldas blog, the court's prima facie findings at this interim stage were:

  • Jurisdiction: Indian courts can hear the case. ANI is based in New Delhi, OpenAI targets Indian users, and storing works on US servers was treated as the final step in a chain that starts with access from India.
  • Training: storing works for training was held, prima facie, to be fair dealing under Section 52(1)(a) of the Copyright Act, 1957, even though the use is commercial.
  • Outputs: no prima facie infringement was shown. The articles ANI cited were published after the training cut-offs of the models named, and the outputs were not found substantially similar to ANI's text. Copyright protects expression, not the underlying facts of news.
  • Balance of convenience: favoured OpenAI, which had already blocked ANI's site from its crawlers.

LiveLaw's commentators noted that the court found the examples likely came from live retrieval of ANI's website rather than memorisation, and that whether such retrieval is "communication to the public" was left undecided because it had not been properly pleaded. The case continues to trial, where memorisation and reproduction can still be argued with evidence. It is an interim ruling, not the final word.

What this means for Indian businesses

For now, an Indian court has signalled that training on publicly available content is not automatically infringement, and that a claimant must show that outputs actually copy protected expression. That is a signal, not a settled rule. A different bench, a full trial or an amendment by Parliament could change it.

Who owns AI-generated work?

United States: no human author, no copyright

In Thaler v. Perlmutter, Stephen Thaler sought to register an image, "A Recent Entrance to Paradise", created by his AI system, listing the AI as author. The Copyright Office refused, the D.C. Circuit affirmed that the Copyright Act requires works to be authored in the first instance by a human being, and on 2 March 2026 the US Supreme Court declined to hear the appeal, as reported by the National Constitution Center. The ruling concerns a work with no claimed human authorship; work made by a person with AI assistance is a different and still developing question.

India: the "person who causes the work to be created"

India's Copyright Act, 1957 already has a clause for computer-generated works. Section 2(d)(vi) says the author of a computer-generated literary, dramatic, musical or artistic work is "the person who causes the work to be created", as quoted by Managing IP, which also notes the view that only a natural person can be an author under the Act. How that clause applies to modern generative AI, and how much human contribution is needed, has not been settled by an Indian court.

The UK test case: Getty Images v Stability AI

Getty Images sued Stability AI in London over the training of Stable Diffusion. On 4 November 2025 the High Court largely ruled for Stability, as reported by The Register. Getty dropped some claims during the trial, and its primary training claim could not succeed in the UK because the model was not trained there. The judge rejected the secondary infringement claim, finding that the model weights are not an infringing copy because the model does not store the copyrighted works. Getty succeeded only on a narrow trade mark point about Getty watermarks appearing in some generated images. The core question of whether training itself is lawful in the UK was left open.

The pattern across these cases

QuestionWhere courts are leaningHow settled
Can an AI be an author?No. US courts require a human author; Indian commentary reads the Act the same wayFairly settled in the US, untested in Indian courts
Is training on public content infringement?Early rulings have not stopped it (ANI interim order, Getty in the UK)Unsettled; interim and jurisdiction-specific
Can outputs infringe?Yes, if they reproduce protected expression; claimants must show actual copyingThe principle is clear, but each case turns on its own facts
Do trade marks matter?Yes. Getty won on watermarks appearing in outputsFact-specific

Practical rules for businesses publishing AI-assisted content

  1. Keep a human in the creative loop. Have people select, arrange, edit and rewrite substantially. Ownership claims are strongest where human creative contribution is real and documented.
  2. Keep records. Save prompts, drafts and edit history for important assets such as brand campaigns, product copy and software. If ownership is ever questioned, evidence of human work matters.
  3. Do not prompt for imitation. Avoid asking a tool to reproduce a named author's text, a specific artist's style for commercial use, a competitor's copy or characters you do not own.
  4. Check outputs for copying. Run important text through a plagiarism check and reverse-image search key visuals. Look out for stray watermarks, logos and recognisable characters.
  5. Read the tool's terms. Who owns outputs, whether the provider offers any IP indemnity, and on what conditions. Terms differ widely across tools and plans.
  6. Verify facts and quotations. AI tools can invent quotes and attribute them to real people or publications, which creates defamation and misattribution risk as well as copyright risk.
  7. Update contracts. Agency, freelancer and vendor contracts should say whether AI may be used, who owns the output and who bears the risk if it infringes.
  8. Protect your own content. If you do not want your site used for AI training, review your robots.txt rules for AI crawlers and your website terms. The Delhi High Court noted that ANI could block crawlers itself.

What to watch next

The ANI trial, any appeal of the interim order, possible amendments to the Copyright Act on text and data mining, and further US and UK rulings will all shape this area. Review your AI content policy at least once a year.

How Cehpoint can help

Our IT consulting service covers strategy, audits and compliance, and can turn the rules above into a written AI-use and content policy that your legal advisers then review. If you run a law practice, our legal industry page explains how we work with Indian law firms on document review and keeping up with changing rules.

Frequently asked questions

What did the Delhi High Court decide in ANI v OpenAI?

On 24 July 2026, Justice Amit Bansal refused ANI's request for an interim injunction. Prima facie, the court held it had jurisdiction, treated storage for training as fair dealing under Section 52(1)(a), and found no proven copying in ChatGPT's outputs. These are interim findings only; the suit continues and memorisation or reproduction can still be argued at trial.

Can I copyright content created with AI in India?

The Copyright Act names the person who causes a computer-generated work to be created as its author, and commentators read the Act as requiring a natural person. No Indian court has yet settled how this applies to generative AI. Your position is strongest where a person contributes real creative selection, editing and arrangement, and you keep records of that work.

Can my business be sued for publishing AI-generated content?

Yes, if the content reproduces someone else's protected expression, uses their trade marks or misattributes quotes, whether or not AI produced it. Courts look at the output itself. Reduce risk by avoiding prompts that imitate specific works, checking important outputs for copying, removing stray watermarks or logos, and verifying every quotation.

Does the AI tool provider protect me if the output infringes?

Sometimes. Some providers offer IP indemnities on certain business plans, usually with conditions such as using built-in filters and not deliberately prompting for infringing content. Free and consumer plans often offer none. Read the current terms of the exact plan you use and record which tool produced which asset.

Need help with this?

Tell us what you are dealing with. We reply with a plain plan, and where it makes sense, a working demo before you commit.