Artificial intelligence (AI), Generative AI, Agentic AI and other AI tools are gradually being adopted by insolvency professionals to enhance productivity. The benefits of AI for claims processing, document summarizing, identification of financial anomalies and evaluation of resolution plans are self-evident, provided that the output of AI tools is put to critical human scrutiny.

Companies are also adopting AI tools with gusto and assimilating them into their operations. This implies that in insolvency cases of the future, resolution professionals (RPs) will need to understand the AI tools used by insolvent companies. These technical skills may be either self-acquired or outsourced.

At the heart of India’s Insolvency and Bankruptcy Code (IBC) is a moratorium on any disruption of services essential or critical to a company. AI tools and the associated cloud infrastructure may qualify as one of the aforesaid services in future cases.

These services must continue if regular payments are made for the period under insolvency. A corporate will have AI services that are billed in advance as well as those that are billed with a lag.

Services billed in advance would include commitments such as reserved cloud capacity and pre-paid usage. Services billed with a lag may include charges for AI model interfaces, ‘tokens’ used by AI platforms, cloud computing beyond monthly quotas and sundry data processing services.

However, what is the course of action if some lagging-bill vendors halt their services for non-payment of invoices predating the insolvency commencement date? Moreover, these vendors may be situated overseas, beyond the jurisdiction of the National Company Law Tribunal (NCLT); the operative legal provisions of cross-border insolvency are still awaited.

The RP is also tasked with duties, primarily taking control of the corporate debtor along with all its assets and determining its financial position. Herein will arise another challenge.

The assets may comprise the corporate debtor’s data and prompts running on a third-party AI platform. How does one segregate the two? And whose asset is the AI response generated? Further, bereft of the platform, for a valuer, what is the value of the proprietary data (unless one assumes the interchangeability of AI models) that would yield similar output given the same inputs?

Another duty of the RP is to comply with every law. We are all aware of Nick Bostrom’s ‘paperclip maximizer’ thought experiment, wherein, in a quest to maximize paperclip production, an AI system would pursue the goal so diligently that it would consume all of Earth’s metal resources.

In similar vein, if an AI agent tasked with quality inspection but trained to maximize profitability clears goods with slight defects, will the RP be liable for such a breach?

Avoidance transactions, those that are impermissible in law and deprive the corporate debtor of assets through shady means, are another issue that calls for attention. However, a deal that seemingly qualifies as an avoidance transaction but was undertaken in the normal course of business may fall outside its realm.

In the case of a corporate debtor whose AI agent executes a transaction, would it be deemed to be one in the normal course of business, even though the same transaction undertaken by a human may have been categorized as probable avoidance? Further, what if its counterparty is also an AI agent? Surely, prima facie, this would be in the normal course of business. But what if agentic models have been trained unfairly by manipulation of model weights?

Thus, an RP must understand how an agent has been trained to act, which may be difficult to fathom even in an ideal scenario as it would require access to training data, structured metadata, input context, reasoning chains, alternatives considered and audit trails.

Liquidation and the distribution waterfall too will face new questions. A liquidation order is a notice under which employees and workers are discharged. Could an AI model vendor, for example, stop supplying services on the argument that an AI agent was part of the firm’s workforce? Also, what will such an agent qualify as under the liquidation waterfall: worker, employee or vendor?

As AI tools proliferate in the world of business, the aforesaid issues and others currently beyond imagination, will come to the fore. Thus, the Insolvency and Bankruptcy Board of India (IBBI) and relevant regulators should come together to study potential scenarios and devise appropriate solutions.

The ministry of corporate affairs is in the process of developing an Integrated Platform for Insolvency Ecosystem (iPIE) that is designed to assimilate all stakeholders of the insolvency ecosystem.

Going a step further, in conjunction with the National Financial Reporting Authority, it could mandate auditors to audit and report on AI tools used by corporates. This information should remain confidential but may be accessible to an RP in the event a company is admitted to insolvency; although the IBC requires a service provider of the corporate debtor to extend cooperation to the RP, a database would be a quicker means of action.

Finally, the Indian legislature and IBBI would do well to hold wide-ranging stakeholder conversations, not only within the insolvency ecosystem but also with technologists who are well versed in the subject of AI adoption by India Inc. If we fail to adapt IBC processes to the AI age, grey areas will provide fertile ground for litigation, with concomitant delays to the process.

The author is an INSOL fellow & interim leader.

Artificial intelligence (AI), Generative AI, Agentic AI and other AI tools are gradually being adopted by insolvency professionals to enhance productivity. The benefits of AI for claims processing, document summarizing, identification of financial anomalies and evaluation of resolution plans are self-evident, provided that the output of AI tools is put to critical human scrutiny.

Companies are also adopting AI tools with gusto and assimilating them into their operations. This implies that in insolvency cases of the future, resolution professionals (RPs) will need to understand the AI tools used by insolvent companies. These technical skills may be either self-acquired or outsourced.

At the heart of India’s Insolvency and Bankruptcy Code (IBC) is a moratorium on any disruption of services essential or critical to a company. AI tools and the associated cloud infrastructure may qualify as one of the aforesaid services in future cases.

These services must continue if regular payments are made for the period under insolvency. A corporate will have AI services that are billed in advance as well as those that are billed with a lag.

Services billed in advance would include commitments such as reserved cloud capacity and pre-paid usage. Services billed with a lag may include charges for AI model interfaces, ‘tokens’ used by AI platforms, cloud computing beyond monthly quotas and sundry data processing services.

However, what is the course of action if some lagging-bill vendors halt their services for non-payment of invoices predating the insolvency commencement date? Moreover, these vendors may be situated overseas, beyond the jurisdiction of the National Company Law Tribunal (NCLT); the operative legal provisions of cross-border insolvency are still awaited.

The RP is also tasked with duties, primarily taking control of the corporate debtor along with all its assets and determining its financial position. Herein will arise another challenge.

The assets may comprise the corporate debtor’s data and prompts running on a third-party AI platform. How does one segregate the two? And whose asset is the AI response generated? Further, bereft of the platform, for a valuer, what is the value of the proprietary data (unless one assumes the interchangeability of AI models) that would yield similar output given the same inputs?

Another duty of the RP is to comply with every law. We are all aware of Nick Bostrom’s ‘paperclip maximizer’ thought experiment, wherein, in a quest to maximize paperclip production, an AI system would pursue the goal so diligently that it would consume all of Earth’s metal resources.

In similar vein, if an AI agent tasked with quality inspection but trained to maximize profitability clears goods with slight defects, will the RP be liable for such a breach?

Avoidance transactions, those that are impermissible in law and deprive the corporate debtor of assets through shady means, are another issue that calls for attention. However, a deal that seemingly qualifies as an avoidance transaction but was undertaken in the normal course of business may fall outside its realm.

In the case of a corporate debtor whose AI agent executes a transaction, would it be deemed to be one in the normal course of business, even though the same transaction undertaken by a human may have been categorized as probable avoidance? Further, what if its counterparty is also an AI agent? Surely, prima facie, this would be in the normal course of business. But what if agentic models have been trained unfairly by manipulation of model weights?

Thus, an RP must understand how an agent has been trained to act, which may be difficult to fathom even in an ideal scenario as it would require access to training data, structured metadata, input context, reasoning chains, alternatives considered and audit trails.

Liquidation and the distribution waterfall too will face new questions. A liquidation order is a notice under which employees and workers are discharged. Could an AI model vendor, for example, stop supplying services on the argument that an AI agent was part of the firm’s workforce? Also, what will such an agent qualify as under the liquidation waterfall: worker, employee or vendor?

As AI tools proliferate in the world of business, the aforesaid issues and others currently beyond imagination, will come to the fore. Thus, the Insolvency and Bankruptcy Board of India (IBBI) and relevant regulators should come together to study potential scenarios and devise appropriate solutions.

The ministry of corporate affairs is in the process of developing an Integrated Platform for Insolvency Ecosystem (iPIE) that is designed to assimilate all stakeholders of the insolvency ecosystem.

Going a step further, in conjunction with the National Financial Reporting Authority, it could mandate auditors to audit and report on AI tools used by corporates. This information should remain confidential but may be accessible to an RP in the event a company is admitted to insolvency; although the IBC requires a service provider of the corporate debtor to extend cooperation to the RP, a database would be a quicker means of action.

Finally, the Indian legislature and IBBI would do well to hold wide-ranging stakeholder conversations, not only within the insolvency ecosystem but also with technologists who are well versed in the subject of AI adoption by India Inc. If we fail to adapt IBC processes to the AI age, grey areas will provide fertile ground for litigation, with concomitant delays to the process.

The author is an INSOL fellow & interim leader.