Reconceptualising the IBC for AI: A Case for Asset Stacks

Raghav Krishna Agnihotri, Madhav Krishna Agnihotri

Sep 1

6 min read

[Raghav and Madhav are students at National Law University Odisha and National Law University Jodhpur, respectively.]

In 2025, the AI industry witnessed one of its first major insolvencies when Builder.ai entered bankruptcy proceedings. This start-up was Builder.ai, founded by Sachin Dev Duggal, and originally valued at USD 1.5 billion. Since then, another 18 Indian AI start-ups, such as Subtl.ai, have also entered bankruptcy.

The scale of the problem is reflected in the extent of India’s AI ecosystem. Investment in Indian Start-up AI ecosystems in Q1 alone was USD 1.48 billion, which accounts for about 38% of all funding into start-ups. However, the stark contrast of a 33% drop in equity funding in Q1 of 2026 follows the unprecedented expansion of AI. As investment slows and pressures intensify, AI enterprises may need to face insolvency, bringing their unique nature within the ambit of the Insolvency and Bankruptcy Code 2016 (IBC).

Yet, the IBC was enacted on the assumption of conventional forms of assets that are capable of value preservation for creditors. AI-native enterprises challenge that assumption, as their primary value lies not in conventional assets but in user-contributed datasets and asset stacks, the legal character of which remains uncertain. This gives rise to a peculiar question: can such non-conventional assets form part of the insolvency estate and, if so, what obligations do resolution professionals (RP) owe in preserving, valuing, and ultimately realising them?

In addressing this question, this piece shall address three issues. First, do AI asset stacks challenge the conventional framework under the IBC? Second, do AI data sets form part of the insolvency estate, and what new obligations for the RP arise given this finding? Thirdly, what is the way forward to form an AI-centric insolvency framework?

AI Asset Stacks Challenging IBC’s Notion of Property

Why are AI asset stacks imperative?

An AI enterprise’s truest reservoir of assets is not computers, inventory, or tangible assets, but in user-contributed data developed over years of interaction with customers. The global AI dataset market is estimated to be USD 4.44 billion.

A single dataset is derived from multiple sources, including customer-generated datasets, publicly available materials, and synthetic data generated by the AI itself. The datasets, weights, and user inputs become the “asset stack”, serving as the foundation for the generative algorithm to develop.

The assumption behind IBC

The basis of the IBC has been repeatedly emphasised as aiming for “maximum utilisation”. To achieve this, the RP is empowered to control and realize assets to satisfy creditors. Implicit in this framework is the simple assumption that a debtor’s value lies in identifiable assets that can be owned, preserved, or transferred.

This assumption is reflected in the architecture of the code. Section 3(27) of the IBC adopts a definition of “property” extending to “movable and immovable property, money, claims, and every definition of property situated inside or outside India.” More significantly, however, Section 18(f) empowers the interim resolution professional to take control and custody only over assets which the corporate debtor has ownership rights. The accompanying Explanation (a) reinforces this ownership-centric approach by excluding assets owned by third parties but merely held by the corporate debtor under contractual arrangements.

The framework sits uneasily with an AI asset stack. Unlike conventional commercial assets, an AI asset stack rarely consists of property owned exclusively by the corporate debtor. Its commercial value derives from the integration of multiple components, including proprietary code, model weights, licensed datasets, user-contributed data, APIs, and third-party software. While some of these components may be owned by the debtor, others exist only through contractual licences or remain the property of third parties.

This fragmented ownership exposes a fundamental limitation. Although an AI asset stack represents one of the most valuable commercial resources, the IBC does not recognise it as a distinct insolvency asset. Instead, its constituent elements fall into different legal categories, such as intellectual property, contractual rights, licensed data, software, and third-party assets. The IBC provides no guidance on how these interdependent components ought to be preserved and realised.

Thus, the absence of a clear legal classification leaves the RP without an adequate framework for dealing with AI asset stacks. While the IBC identifies the assets owned by the debtor, it does not address how an integrated technological asset, which depends on both owned and licensed components, should be preserved or transferred. As a result, the preservation and realisation of AI asset stacks are questions that the existing framework is not equipped to answer.

AI Asset Stacks being part of the IBC Insolvency Estate?

Why AI asset stacks should form part of the insolvency estate

Although the IBC does not expressly recognise AI asset stacks as assets, their exclusion would defeat the Code’s objective of maximising utilisation for the corporate debtor. The Supreme Court in Victory Iron Works Limited v. Jitendra Lohia & Another and in State Bank of India v . Union of India recognised that the IBC is not concerned with mere proprietary ownership but with commercial interests capable of preservation and transfer.

The relevant enquiry is therefore whether the corporate debtor possesses a transferable commercial interest capable of preservation and realisation during the insolvency resolution process. Training datasets may incorporate licensed material and user-contributed data, while cloud infrastructure, APIs, and software frequently operate through contractual arrangements. Nevertheless, the debtor possesses an integrated bundle of proprietary and contractual rights that enables it to develop, deploy, and commercialise the AI system.

Applying this reasoning, it is this integrated commercial interest, rather than absolute ownership of each underlying component, that should be regarded as forming part of the insolvency estate.

Reconceptualising RP’s duty

Once the AI asset stack is recognised as the relevant insolvency asset, we find another problem: the RP’s statutory obligation to preserve assets in a manner that retains their commercial value. This presents a challenge because, unlike conventional assets such as factories or machinery, the value of an AI asset stack depends upon continuous updates, maintenance, and the quality and integrity of the underlying data. Physical control over servers or digital infrastructure alone cannot preserve the asset in any meaningful sense.

Accordingly, the statutory obligation to take “custody and control” must be understood functionally rather than physically. The RP must preserve the conditions that enable the AI Asset Stack to remain commercially exploitable, including maintaining API access, complying with licence terms, safeguarding proprietary datasets, preserving records, and ensuring compliance with applicable data protection obligations. Failing to do so risks eroding the very asset whose value the insolvency process seeks to retain.

Towards an AI-Native Insolvency Framework

While no jurisdiction has yet developed a comprehensive insolvency framework for AI asset stacks, comparative practice demonstrates an emerging recognition that AI-driven assets require treatment beyond conventional norms.

In the United States, bankruptcy proceedings have recognised software, proprietary databases, customer contracts, and other intangible assets as valuable components of a debtor’s business that may be transferred as part of a going concern.
Similarly, the European Union’s AI Act does not address insolvency; it adopts a lifecycle approach to AI governance, recognising that AI systems derive value from the continued management of data, models, and compliance obligations rather than from isolated assets. This aligns with the ideal of the IBC.

Collectively, these developments indicate that AI enterprises require a framework that reflects their unique commercial reality, in the form of a coherent insolvency theory.

Coming to What India Should Do

India should adopt an interpretation of the IBC that recognises AI asset stacks as a category of insolvency assets. Further, instead of focusing on who owns each dataset or AI model, the RP should identify the asset stack as the bundle of commercial rights that gives the business its value. The RP would preserve the technical conditions necessary to maintain the stack’s value, transferring only those rights that can lawfully be transferred, subject to applicable data protection laws.

Such an approach advances the IBC’s objective of value maximisation while respecting existing principles of property, contract, and privacy law. Most importantly, it enables the insolvency framework to deal effectively with AI-driven businesses, whose value lies in an integrated AI Asset Stack rather than in separate digital assets.

Conclusion

The IBC was designed on the assumption that a company’s value lies in assets that can be easily owned and transferred; however, emerging trends consistently challenge these ideals. AI companies challenge this assumption because their value lies in AI asset stacks made up of datasets, models, licences, and other connected digital resources.

This article argues that, instead of focusing on who owns each individual component, the IBC should recognise the corporate debtor’s commercial interest in the AI asset stack. Such an approach is consistent with the IBC's objective of maximising value while remaining compatible with existing principles of property, contract, and data protection law.

As AI becomes a larger part of India’s economy, recognising AI asset stacks under the IBC will help ensure that valuable AI businesses can be effectively preserved and resolved during insolvency.