How Generative AI Is Shattering The Core Of Indian IP Agreements With An Impossible Promise

Any creator, developer, or artist who has ever licensed their work has seen a line in their contract which is both foundational and forgettable. This line is deep in the legalese, yet forms the bedrock upon which the entire agreement stands. It reads something like this:

"The AUTHOR hereby declares that the said WORK is an original WORK and that the AUTHOR is the owner of the copyright therein and that the WORK is in no way whatsoever a violation of any existing copyright."

A simple promise which is a warranty of originality and serves as indemnity against infringement. This clause in licensing and IP agreements has for decades been the handshake which allows intellectual property to move from an individual's mind to the global marketplace. It is the seed behind any Intellectual Property, the promise of that very 'intellect.' The creator warrants they created it, and the publisher or licensee accepts the work, secure in that knowledge.

Then came the era of Generative AI.

The usage of Generative AI has already become so ingrained into our daily lives that suddenly, any creator who drafts code, writes prose, or designs images is using AI, and these creators can no longer make that simple promise in good faith. AI systems have largely been trained on terabytes of data which the creators have zero insight into, they cannot audit the AI's process. There is no guarantee that an elegant line of code or a perfect paragraph wasn't memorised and reproduced from someone else's protected work which the AI was trained upon.

To sign that clause, knowing your work was co-created with this 'black box,' is to make an impossible promise.

The Breakdown of IP Risk and Contractual Reality

To grasp the severity of this impossible promise plaguing the IP contracts in circulation, we must start by distinguishing between a copyright infringement claim and a breach of warranty.

Infringement is a tort. A third party needs to launch a legal challenge, then prove substantial similarity, and then navigate the complexities of litigation usually under the Copyright Act of 1957. It is a costly, time-consuming endeavour and its outcome is never certain.

A breach of warranty, on the other hand, is a much simpler affair. A breach of warranty is a violation of the contract itself. The clause in question is purely commercial and exists to shift the entire legal and financial risk of an IP claim from the publisher or platform onto the shoulders of the creator as it should be.

When a creator warrants originality, they indemnify the other party, in this case the person taking on the IP of the creator. In case a lawsuit is filed, the publisher doesn't need to wait for a court to rule on the merits of the infringement as they can simply point to the contract, declare a material breach of warranty, terminate the agreement, and sue the creator for damages.

Unfortunately, with Generative AI this mechanism is completely shattered. For the first time, millions of creators are producing work where they cannot stand behind its originality. If a creator even uses AI for simple transitions then the risk is formed. This is not a theoretical argument about AI but a clear and present crisis happening in this very moment, with every signature on most IP licensing agreements in India and perhaps even across the world.

The Publisher's Gamble and Current Indemnity Structure

So, a creator's promise is broken. This brings the question of why a multi-billion-rupee platform or publisher care? After all, the risk has been transferred on paper to the creator, and there still exists the constitutional shield of Safe Harbour.

For over a decade, the safe harbour provisions of Section 79 of the Information Technology Act, 2000 have been the bedrock of India's digital economy. We have seen in cases like MySpace v. Super Cassettes that an intermediary cannot be liable for infringing content since it lacks "actual knowledge" of the illegal activity.

This distinction between a passive host and an active publisher is the most important aspect of this argument. A platform that pays for and publishes a script co-created with AI under a contract is anactive participant and not a passive host. The platforms have "actual knowledge" of the systemic and unavoidable risk that the AI and the author relying on it for anything at all may produce infringing work. Proceeding in the face of this known risk is arguably a form of wilful blindness that could dissolve the protections of Section 79.

This is further where the chain of liability becomes a complete and catastrophic failure. Let us imagine a scriptwriter who utilizes an AI tool for a new web series. The writer signs a contract with a production house and the production house further signs a larger deal with a major OTT platform. The show is a smashing hit, and that's when a novelist discovers entire passages were lifted from their book due to even the slightest use of AI.

Who gets the first legal notice? The deep-pocketed platform. The platform then turns to the production house for indemnification, that in turn sues the writer for breach of warranty. This system collapses in practice as a warranty from an individual freelancer with limited assets is a paper shield against a multi-million-rupee claim and the platform and publisher still have to get enrolled in a legal battle but also suffer major reputational damages

The risk they thought they had contracted away comes back amplified. The legal frameworks which were designed to manage IP risk become the source of an even greater, unmanageable one. This is the nature of the gamble that publisher's take on with each contract they sign today while Generative AI colours the ranks of artists all across the world.

The Need for a New Promise in Contracts

The natural instinct of the legal landscape is to wait. Wait for Parliament to amend the Copyright Act. Wait for the Supreme Court to hand down a landmark ruling. Wait for clarity. Commerce, unfortunately, does not wait.

The industry needs a solution in the next five months and not five years. Waiting for a legislative answer would take too long hence we have to turn to a contractual solution for this major issue.

We need to throw out the old, broken promise and forge a new one. For this an addendum to existing agreements should be proposed that confronts the problem head-on. A tweak which would lead to a reimagining of risk allocation for the age of AI.

The architecture of such an addendum would then be based upon the following major aspects:

1. Radical Transparency.
   The foundation of the addendum would be a shift from warranty to disclosure. The addendum can require the creator to declare upfront, whether Generative AI was used in the creation of the work. It would specify which tools were used and, to a reasonable extent, for what purpose—ideation, first drafts, editing, code generation.
2. Modified Warranty
   The second step after the disclosure would be to add a new, more precise set of promises that the creator can actually make. Instead of warranting the AI's output, the creator warrants their process. The creator can promise that their prompts and inputs were original, that they did not intentionally use the AI to replicate any existing work, and that they have applied substantial human skill, creativity, and judgment to curate, edit, and arrange the final output hence tying the promise to a human element.
3. The Indemnity Shift.
   The last step is the proper allocation of risk. The creator can agree to indemnify the publisher against any third-party claims arising specifically from the AI-generated portions of the work as declared. The publisher hence accepts the content with full knowledge of its origins, and in return, receives a clear and enforceable contractual protection for that risk. This indemnity can be negotiated or perhaps capped at a certain monetary value or linked to the advance paid to make it more than just a paper shield.

This approach then provides a workable framework for the publisher and the creator especially by adding clarity. A publisher can make a commercial decision based on known facts which would allow them to adjust advances, increase scrutiny, or even reject works that are too heavily reliant on AI, and for the honest creator, it allows for innovation with accountability.

Conclusion

The law, in its deliberate way, will eventually catch up. The courts in cases like the one brought by ANI against OpenAI in Delhi will draw new lines. Our Parliament may one day amend the Copyright Act of 1957. But the digital economy does not have the luxury of waiting. It operates in the here and now, on the strength of the contracts signed today.

The future of legal practice, especially at the volatile intersection of intellectual property and technology, will not be defined by those who can best litigate what is left of old agreements but will belong to those who can engineer new ones. We must build the frameworks that allow innovation to flourish without a vacuum but with a commercially viable framework.

References:

• The Copyright Act 1957
• The Indian Contract Act 1872
• The Information Technology Act 2000
• ANI Media Pvt Ltd v OpenAI, LLC & Anr CS(COMM) 8/2024 (Delhi High Court)
• MySpace Inc v Super Cassettes Industries Ltd FAO (OS) 540/2013 (Delhi High Court, 23 December 2016)
• Carlini N and others, 'Extracting Training Data from Large Language Models' in 30th USENIX Security Symposium (USENIX Association 2021)
• Chaudhary P and Anand A, 'AI And Copyright Law In India: Exploring The Contours Of A New Legal Challenge'
• Federation of Indian Chambers of Commerce & Industry (FICCI), 'The Indian Publishing Industry: An Overview'

The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.