MODERNISING THE IBC FRAMEWORK: DETECTION OF PREFERENTIAL AND FRAUDULENT TRANSACTIONS

- by Ujjawal Priyadarshi and Tanushree Singhal, 3rd year students at Chanakya National Law University, Patna

Introduction

The Insolvency and Bankruptcy Code, 2016, was framed as a relatively limited time-span system that is designed to rescue enterprises which are in a state of financial distress, as well as to extract the maximum value of the assets by the creditors. The Corporate Insolvency Resolution Process (CIRP) has tough time schedules, and creditors control procedures that help the speedy resolution and ensure the integrity of assets. Empirical data, however, support the widespread delays; the average time of CIRPs resulting in the resolution plans is now 581 days, which is significantly more than the stipulated timeframes, and liquidations are even more subject to delay.

One of the challenges which is very critical related to the issue of manual identification of the Preferential, Undervalued, Extortionate Credit Transactions and Fraudulent (PUEF) under the IBC. Resolution   professionals must review two-year retrospective transaction records, identify beneficiaries, and flag anomalies before recommending to the adjudicating authority. This manual process faces a backlog exceeding ₹2.8 lakh crore. To cope with these predicaments, the government is pushing the forced application of Artificial Intelligence (AI) by resolution professionals to automate identification, further transparency, and to handle the abundance of financial information. However, AI installations can also have risk factors like algorithmic bias, and has the potential to legalise discriminatory patterns on lending and insolvency adjudication. Therefore, the strategy of compromising efficiency potential with equality and responsibility continues to be central to the Indian insolvency ecosystem.

The State of the Manual PUEF Detection Framework and Its Shortcomings.

Section 43, 45, 50, and 66 of the IBC outline the avoidance and fraud offences but they are rarely detected because of the operational limitations. Preference claims become worse when the manual reviews are not able to properly timestamp and trace overlay transfers between affiliates. Underpriced sales are still not apparent in situations where documentation of the cases lacks market comparators and forensic values that are scalable. Unrealistically burdensome credit conditions and fraudulent transfers are hidden in the small print of contracts, side deals, shell companies, and cross-border transactions that humans cannot observe, thus making regulators promote the integration of AI that can establish relationships between entities, compare deals, and identify abnormal behaviour.

In the landmark case of Anuj Jain IRP for Jaypee Infratech Limited Vs. Axis Bank Limited etc. the Supreme Court ordered the volumetric and gravimetric analysis of the preference deals; nevertheless, foremost resolution professionals are often faced with tight deadlines (75 -135 days), which are often too hard to manage. The Regulation 35A also demands such applications to be submitted to the tribunal by the 130th day, thus putting extra burden on practitioners. The decision by courts has made it clear that avoidance applications are not part of the CIRP framework and these applications can continue after the framework is ended, thus increasing the burden of procedures. The cases of Byju’s and WazirX are highly sensitive cases that demonstrate that advanced transfers and cyber-attacks demonstrate the weakness of manual detection. There was an illegal transfer of funds amounting to $533 million in the case of Byju’s, which was executed when it was undergoing insolvency undetected by the lenders and in the case of WazirX the interplay of issues concerning the recovery of assets. The cumulative effect of these incidents is the need to seek technological interventions that can overcome human bottlenecks and time loss in lawsuits.

Artificial intelligence promises insolvency resolution.

The Insolvency and Bankruptcy Board of India (IBBI) has suggested the mandatory use of AI amongst resolution specialists, and the purpose of finding fraud, PUEF transactions, and to improve the forensic analysis in view of the data volume and complexity. AI solutions have a potential to automate documentation, detection of anomalies and reporting to stakeholders. Empirical studies show machine-learning models like logistic regression and random forest classifiers achieving over 74% accuracy in detecting fraudulent filings. Ensemble methods, including decision trees, support vector machines (SVMs), and neural networks (NNs), reach up to 95% on large datasets.

The practice of AI shows how possible its usage can be. The AI Strategy of the United Kingdom, Finland based KOSTI insolvency platform, and the AI based bankruptcy portal of Colombia have provided impactful efficiencies and better case management with the automation of non-discretionary services, but not human supervision. The Supreme Court of India has recognised the potentials of AI to supplement the insolvency process in India, which has been time-bound by forecasting the insolvency risk and allowing the case to proceed.

Striking a Balance between AI Efficiency and Creditor and Debtor Rights.

AI systems are expected to replicate the biases within their training data, which may result in discriminatory lending patterns and unaccountable and unintelligent decisions that unduly impact stakeholders. Evidence based on empirical data has shown that minority borrowers are denied loans at higher rates and that they have prolonged processing times compared to similar non-minority borrowers, proving that historical bias in datasets can be converted into harmful automated applications. In this regard, sections 8, 10, and 11 of the DPDP Act, 2023 act as a general protection by ensuring that accuracy and accountability are upheld in the use of data, and that, affected individuals have the ability to request correction and redress in the event that the insolvency or credit-related decisions are based on biased or incomplete data processed through AI. False-positives decision making occurs where AI incorrectly recognises legitimate behaviour as fraudulent, and this issue makes one concerned about the contextual sensitivity of AI. These collapses pose risks on the confidence of the people in the insolvency system.

The Digital Personal Data Protection Act of 2023 introduced by India presents privacy and consent considerations to the AI implementation within the company in the insolvency process, and this makes it hard to reconcile the necessity to minimize data and the needs of AI. The Jet Airways case was a manifestation of collision between the requirements of the CIRP data and privacy law. Courts are faced with challenges of admitting AI-generated evidence because of the technicality and authentication challenges. There are growing expectation and compulsion of rights to explanations of automated decisions in global jurisdictions through the General Data Protection Regulation (GDPR) requirements and the emerging EU AI Act. In this respect, India should come up with algorithmic accountability mechanisms, including impact assessment, mandates of transparency, and judicial recommendations on admissibility of AI evidence in order to assure fairness.

Case Study: Obstacles and Regulation Implications of Implementation

The general obstacles to broad AI use include institutional factors, such as training needs; the Training-of-Trainers program of the IBBI is aimed at developing insolvency specialists. The shortage of skills and change to technological tools is a hitch. Progress should be accompanied by special AI literacy, as well as transversal soft skills. Banking institutions find it hard to manage infrastructural expenses.

The recent changes in regulations by the IBBI incorporate AI tools into the current framework including Regulation 35A into the disclosure requirement system and digital document practices. National Company Law Tribunal (NCLT) is required to be ready to review AI-based evidence with procedural rigor. Regulatory advice on how to validate the tools, weight of evidence, as well as transparency is necessary. It would be advisable to use gradual rollout plans that include pilot projects, risk management models, stakeholder partnerships, and adherence management, so that the resolution professionals can gradually adopt AI tools into the known CIRP processes and meet the deadlines set out in statutes to review transactions. Further, an effective safe harbour for resolution professionals involved in the implementation of validated AI tools could also reduce evidentiary uncertainty and risk of the process when presented before the NCLT. There should be the enforcement of data-security standard in line with the requirements of DPDP Act. An IBBI AI department should be specialised in certifying tools and this division should carry out compliance auditing.

Conclusion

The innovations of the Indian insolvency system should focus on the replacement of manual insolvency fraud detection with AI-based one. Paper-based techniques are ineffective when the level of complexity and data increase. AI provides dynamic detection and decision-making opportunities which are scalable. However, a rational solution that will see to it that AI performance does not undermine the rights of the creditors or the rights of the debtors is urgent. Mitigation of bias and discrimination risks should be done through transparency, accountability, and ethical using AI. India requires stakeholders, such as policy makers, adjudicators, resolution professionals, and technologies, to cooperate to establish transparent, fair and efficient AI implementations that are in line with legal and societal values of India. Hopefully, the establishment of an insolvency system that is technologically up-to-date and fair will be possible with the holistic protection considered, including data protection, judicial standards, and safe harbours, which will give India a chance to build trust among the people.