<?xml version="1.0" encoding="utf-8"?>
<!DOCTYPE article PUBLIC "-//NLM//DTD JATS (Z39.96) Journal Publishing DTD v1.2 20190208//EN" "https://jats.nlm.nih.gov/publishing/1.2/JATS-journalpublishing1.dtd">
<article xmlns:xlink="http://www.w3.org/1999/xlink" article-type="research-article" xml:lang="en">
  <front>
    <journal-meta>
      <journal-id journal-id-type="publisher-id">IISTJ</journal-id>
      <journal-id journal-id-type="issn">3139-1478</journal-id>
      <journal-title-group>
        <journal-title>International Innovations &amp; Scholarly Trends Journal</journal-title>
        <abbrev-journal-title>IISTJ</abbrev-journal-title>
      </journal-title-group>
      <issn pub-type="epub">3139-1478</issn>
      <issn pub-type="ppub">3139-9886</issn>
      <publisher>
        <publisher-name>IISTJ</publisher-name>
      </publisher>
    </journal-meta>
    <article-meta>
      <article-id pub-id-type="publisher-id">122020</article-id>
      <article-id pub-id-type="doi">10.5281/zenodo.24299273</article-id>
      <article-categories>
        <subj-group subj-group-type="heading">
          <subject>Law and Legal Studies</subject>
        </subj-group>
      </article-categories>
      <title-group>
        <article-title>Can Artificial Intelligence Be an “Officer in Default”? Rethinking Corporate Liability under the Companies Act, 2013</article-title>
      </title-group>
      <contrib-group>
        <contrib contrib-type="author">
          <name>
            <surname>Baru</surname>
            <given-names>Aastha</given-names>
          </name>
          <email>aastha.baru1705@gmail.com</email>
          <aff>Svkm</aff>
        </contrib>
      </contrib-group>
      <pub-date pub-type="epub">
        <day>17</day>
        <month>11</month>
        <year>2026</year>
      </pub-date>
      <volume>2</volume>
      <issue>11</issue>
      <fpage>321</fpage>
      <lpage>328</lpage>
      <permissions>
        <copyright-statement>© 2026 The Author(s). Published by IISTJ.</copyright-statement>
        <copyright-year>2026</copyright-year>
        <license license-type="open-access" xlink:href="https://creativecommons.org/licenses/by/4.0/">
          <license-p>This is an open-access article distributed under the terms of the Creative Commons Attribution 4.0 International License.</license-p>
        </license>
      </permissions>
      <self-uri content-type="landing-page" xlink:href="https://www.iistj.org/article.php?paperid=122020"/>
      <abstract>
        <p>According to the Companies Act, 2013, liability in a corporation is not introduced to the corporation as an abstract person rather it is imposed on the individuals, described as “officers who are in default”, through the definitions of the word “officer” in Section 2(59) and “officer who is in default” as defined in Section 2(60). That means a human agent is assumed to be knowledgeable and exercising autonomy in making decisions, giving consent to comply with statutory obligations, being aware of a default through Board meetings and possessing sufficient mens rea that would entail imprisonment or payment of fines, as stated in Sections 447 to 451 of the Act. This assumption is gradually being questioned with the introduction of artificial intelligence (AI) into compliance, reporting, risk assessment, and making decisions at the Board level, which finds its roots in Lennard’s Carrying Co. v. Asiatic Petroleum Co.&#13;
If a company acts in a certain way due to an AI system rather than a named officer, it will strain the whole statutory framework for attributing wrongdoing. It would be difficult to call a system that does not have legal personality, mens rea, or capacity to give written consent an “officer in default.” In this paper, I analyze each statutory element of Section 2(60)—legal personality, prior written consent, deemed knowledge based on the Board proceeding, and the discretionary relief under Section 463 for officers that behaved honestly and reasonably—and demonstrate that none of these elements can be applied to this kind of non-sentient AI system without tearing apart the protective measures that Parliament intentionally included in this legislative provision.&#13;
It also analyzes the various responses to the idea of AI accountability, namely the European Parliament’s 2017 suggestion to create an entity called ‘electronic person,’ the accountability framework established in the EU Artificial Intelligence Act, the notion of ‘shadow director’ in England and the Indian approach to the matter (with reference to SEBI’s framework on trading through algorithms). It then proceeds to define what the author refers to as ‘the accountability gap,’ which refers to the diffuse responsibility and also to the problem of ‘liability laundering.’ To address this gap, the author comes up with reforms proposed based on the current legislation.</p>
      </abstract>
    </article-meta>
  </front>
  <body>
    <sec id="sec1" specific-use="heading-1" sec-type="intro">
      <title>Introduction</title>
      <p>The principle of corporate liability does not consist in penalizing the concept or the term ‘company’ as such but it holds human beings liable who are called officers of the company and hold responsibility for the problems faced by the company. Thus, companies can act and function through human beings and the Companies Act of 2013 introduces an interesting mechanism of attribution whereby certain individuals can be defined as officers who are morally and legally responsible for any wrongdoing committed by the company. The basic idea behind this framework is that even if institutions are not always run by identifiable people, it will eventually be possible to trace those institutions' decisions back to identifiable leaders. In other words, corporate governance has traditionally been viewed as the prerogative of human beings, but this view is undergoing a shift as more firms use artificial intelligence systems instead of people to make decisions such as approve transactions, calculate risk levels or drawn up compliance documents. In light of this transformation, the question arises as to whether the AI system can be regarded as a person responsible for the wrong done, and if not, who will be held liable in case of errors done by the system. The basic idea behind this framework is that even if institutions are not always run by identifiable people, it will eventually be possible to trace those institutions' decisions back to identifiable leaders. In other words, corporate governance has traditionally been viewed as the prerogative of human beings, but this view is undergoing a shift as more firms use artificial intelligence systems instead of people to make decisions such as approve transactions, calculate risk levels or drawn up compliance documents. In light of this transformation, the question arises as to whether the AI system can be regarded as a person responsible for the wrong done, and if not, who will be held liable in case of errors done by the system.</p>
    </sec>
    <sec id="sec2" specific-use="heading-1">
      <title>THE STATUTORY FRAMEWORK: “OFFICER IN DEFAULT” UNDER THE COMPANIES ACT, 2013</title>
      <p>Definitions under Section 2 (59) and 2 (60) Section 2 (59) defines “officer” includes any director, manager or key managerial personnel or any person in accordance with whose directions or instructions the Board of Directors or any one or more of the directors is or are accustomed to act. 1 This broad definition is designed to prevent evasion of liability through indirect control. Section 2 (60) defines “officer who is in default” and it narrows down the class for the purpose of penal and quasi penal liability, providing an exhaustive list of categories of officer who is in default” 2 These include “the whole-time director, each key managerial person including managing director, chief executive officer, chief financial officer and company secretary, and any nominated directors in their absence”3 There are two important characteristics relevant to the present discussion. Firstly, some terms are used as formal appointments or designations and thus refer to human entities specified in Sections 2(94) and 203 of the Act. 4Secondly, some terms are used in terms of consent or knowledge; thus, a director responsible for compliance is liable to the authorities provided he/she has given his/her consent in writing beforehand; a director becomes an officer violating the law as long as a violation took place with his/her knowledge attained by the Board and he/she never opposed it.5 Section 149(12) also limits the liability of independent and specified non-executive directors regarding matters that they are aware of and to which they have given consent, expressed intent, or failed to act with diligence. 6Section 463 gives discretionary power to courts to pardon any official “who has acted honestly and with reason associated with the situation.”7 The Doctrine of Identification and Attribution Section 2(60) replicates the principle of the common law doctrine of corporate attribution, ‘directing mind and will’, considered in Lennard’s Carrying Co. v. Asiatic Petroleum Co8, but later on clarified by the court in Tesco Supermarkets Ltd. v. Nattrass to signify those who operate a business in reality. 9In Indian jurisprudence, this principle is also in effect. Accordingly, the court found that a company could be hit with corporate liability. In other words, the common principle of attribution is at play here, meaning that if the human mind (directors) knows or intends a crime, this leads to the attribution of liability against the corporation.10 The main idea behind corporate liability is attribution, referring to the transfer of liability from an organization to an individual. Thus, there must be some identifiable human ‘mind’ at the moment of decision-making. However, if it is not a human mind that is responsible for this decision but rather a computer-based model, then there is nobody to be held accountable.</p>
    </sec>
    <sec id="sec3" specific-use="heading-1">
      <title>AI IN CORPORATE GOVERNANCE AND DECISION MAKING</title>
      <p>The progression of AI systems is from simply assisting in making decisions to making decisions on their own1. At one end of the spectrum lie such AI functions as generating summaries of board meetings, spotting discrepancies in financial statements and estimating the likelihood of litigation. These are purely advisory functions, with the final word belonging to the human executive. 2At the other end, widely used by regulated finance companies and large organizations, AI performs tasks that were previously within the remit of senior management, such as approving related party transactions based on criteria established by AI, or issuing internal compliance obligations by comparing ledgers without any human involvement.3</p>
    </sec>
    <sec id="sec4" specific-use="heading-1">
      <title>CAN AI SATISFY THE STATUTORY TEST FOR “OFFICER IN DEFAULT”</title>
      <p>A. The threshold requirement of a legal personality: Each category in Section 2 (60) implies the existence of a person who holds a given position and who can be designated or prosecuted under Sections 447 to 451 of the Act, with imprisonment and pecuniary punishment being possible penalties. 1An AI system cannot be considered as a person according to Indian law regardless of its development level. Moreover, it does not hold any directorship, since Section 149(1) distinguishes individual persons only. 2Therefore, an AI system does not cover any definition from Section 2 (60)it cannot be classified as a whole-time director, a top manager or a person with some accountability under restrictions.3 B. Mens Rea, Written Consent and the Absence of Non- Human Agency: In concession of the idea that there may be flexible interpretations of the definitional parameters, the operational safeguards contained in Section 2(60) render the extension impossible in and of themselves.4 The consent limb demands that this consent be expressed “in writing beforehand,” which can only take place through an agent that understands and agrees to take up responsibility that is something that cannot be achieved through using an AI system within the specific meaning of the term established by the law. The knowledge limb indicates that the person has to be aware of the breach through “the proceedings of the Board” and has not raised objections. Examples of Section 463’s discretionary relief demonstrate that the essence of the limits is in the requirement of whether the officer “acted honestly and reasonably,” which is not applicable to a non-sentient system. In every case, it can be understood that the safeguards act here not as incidental ones but rather as distinguishing features of the notion of officer in default: they are exactly what allows punishing the person as such. Applying this concept to AI will imply throwing out the safeguards altogether or keeping them but rendering them meaningless. 5 C. Attribution Through the Human Proxy: Where Responsibility Must Instead Fall If AI cannot itself be found liable for default, the next question is whether established categories can still encompass those human beings who use or rely upon AI systems, as suggested by the broad language of ‘accustomed to act’ in Section 2(59).6A Board that acts solely on AI's recommendation could thus be viewed as a Board that acts on advice provided by ‘any individual’ as mentioned in the Section 2(59) provided that in extreme cases, the vendor, developer, or in-house team involved in the creation of AI system might fall into the category of ‘any person’ given the broad interpretation of the law. The advantage of this approach is that it allows identification of a human (or any other corporation) that could be held liable without amendment of the law. Yet it is not without issues: Section 2(59) was created with informal human influence in mind, such as promoters, financiers, and controlling shareholders, and the courts have always been careful to restrict its applications.7 Relying on strained interpretation, rather than clear statutory language, to resolve a genuinely novel problem risks inconsistent enforcement and offends the rule-of-law preference for fair notice, particularly given that Sections 447–451 attach criminal consequences.8</p>
    </sec>
    <sec id="sec5" specific-use="heading-1">
      <title>INTERNATIONAL PERSPECTIVE</title>
      <p>In 2017, the European Parliament’s Committee on Legal Affairs proposed that the European Union should have a separate legal classification for robots, which would include a form of ‘electronic personhood’ for advanced automated machines. This proposal was condemned by more than 150 scholars of AI, robotics and law who signed an open letter claiming that electronic personhood is redundant in legal terms and can be exploited to relieve the producers of machines from liability and put the blame on a fictitious legal concept unable to be held accountable. 1The proposal was not passed into law and the Artificial Intelligence Act passed by the European Union afterwards introduced a human accountability rather than machine personhood based model of legislation, which made the parties involved in the process of creation and implementation of new machine technologies liable for them. 2This historical episode shows that the field which is most open to granting a legal status to AI came to conclusion that the future of the legislation in this area lies in human and corporation responsibility rather than machine personhood model, which is a point that this article supports in relation to Indian company law.3</p>
      <p>An analogy can be drawn from English law as regards to the ‘shadow director’ set out in Section 251 of the Companies Act 2006, which refers to a party whose direction or instruction is taken by the directors of a company in carrying out their work. 4While there has been some discussion about the applicability of the term ‘shadow director’ for algorithmic systems, it is suggested that one should still exercise some caution when putting the concept into practice. Instead of extending the definition, the view taken in the paper is that the idea of shadow director has been created to refer to human agents involved in the operation of a corporate structure.5</p>
      <p>In India, The Securities and Exchange Board of India’s framework for algorithmic trading fixes responsibility for an algorithm’s market conduct on the registered stockbroker and algorithm provider, not on the algorithm itself.6 Similarly, the safe-harbour architecture of Section 79 of the Information Technology Act, 2000 conditions an intermediary’s protection from liability on human due diligence, again refusing to treat the automated system as an autonomous bearer of obligation.7 These sectoral precedents support extending the same human-accountability logic to Section 2(60), rather than importing an untested concept of machine liability into company law.8</p>
    </sec>
    <sec id="sec6" specific-use="heading-1">
      <title>THE ACCOUNTABILITY GAP AND DOCTRINAL CHANGES</title>
      <p>Just because it has been established that AI cannot be considered to be ‘in default’ does not mean that all the obstacles that led us to asking this question have been resolved. There are still three issues: first, the so-called ‘many hands’ problem means that when the output of an AI system is being communicated through many human crossings, it becomes impossible to identify a person who meets the necessary conditions established by Section 2(60). 1Second, there is the issue of the lack of proper documentation since AI systems in general (especially deep-learning systems) do not create any reasoning chain, which makes it impossible to know after the fact whether or not the reliance of the director on the NJ was justified under Section 463 or whether the director had sufficient knowledge as stipulated in Section 2(60)(v).2 Third is what this paper terms ‘liability laundering’: the use of an AI system, whether deliberately or through institutional complacency, to create a plausible narrative of good-faith reliance on a supposedly neutral technical process, thereby insulating the true decision-maker from the consequences the Act intends to impose. The existing regulation of auditors and secretaries has not yet mandated businesses to provide evidence of the impact of their artificial intelligence systems on specific board decisions, compliance certificate statements, and related parties’ approvals. This creates a gap in documentation where Sections 2(60) and 463 demand the utmost information.3</p>
    </sec>
    <sec id="sec7" specific-use="heading-1">
      <title>PROPOSALS FOR REFORM</title>
      <p>The Act should be modified to include a new deeming provision essentially a new sub-clause in Section 2(60) that considers the officer who authorized or approved the implementation of an AI system for a certain compliance, financial reporting, or related-party function as an officer in violation in reference to violations linked to that AI system's output, unless he or she is able to demonstrate that there was really human verification. This shifts the burden of proof in a similar way to what is done in the consent-based component of Section 2(60)(iv) the same way that a person who consents in writing to take responsibility for compliance accepts the risk that arises from it, a person who permits the use of an AI system for performing a statutory function should also take responsibility for the failures of this system, whereby a reasonable defence is available.1 As per Section 134(3), along with Section 134(5), the directors of the company need to furnish a statement indicating that proper internal financial controls have been adopted. In line with this requirement, the rule-making provision under Section 134(3)(a) must mandate the disclosures regarding the AI systems being used in the course of finance reporting/compliance, the extent to which the decision outputs have been reviewed by humans, and the name of the person responsible for such review. 2The necessity for this type of disclosure emerges as it would convert the unnoticed technical process into a documented decision-making trail, which can serve as the basis for the determination prescribed by Section 2(60). This would address the important opacity issue discussed above in Part VI.3 Sections 179 and 188 imply that the Board or a committee uses independent judgment when delegating powers or approving transactions involving related parties. It is necessary to amend or clarify these rules through regulations from the Ministry of Corporate Affairs so that any delegation of authority under Section 179 to any AI-supported process requires a named officer to supervise it. Similarly, any related-party transactions which are either created or screened by an AI system must be indicated the Audit Committee in accordance with Section 177.4</p>
    </sec>
    <sec id="sec8" specific-use="heading-1" sec-type="conclusions">
      <title>CONCLUSION</title>
      <p>The term “officer in default,” found under Sections 2(59) and 2(60) of the Companies Act of 2013, is based on the understanding which has governed Indian company law for more than a century: that every corporate crime is committed with the assistance of some identifiable entity that is responsible for it.1 Artificial intelligence technology does not refute this presumption but puts pressure on the methods of verification of the presumption. Under any interpretation of the statute, an AI system cannot be considered an “officer in default” since it has no legal capacity, is incapable of granting consent in writing, does not possess the necessary knowledge and intentions pursuant to the Act and unable to carry out the punishment defined by the Act. Despite that, it would be wrong to treat the default taking place with the use of artificial intelligence technology as unaccountable or to let impersonal corporate processes mitigate individual accountability. The more effective option is to reinforce rather than reject the human accountability approach that is represented by Section 2(60), and which is supported by an analysis of experience from across the globe, as well as the case of the regulation of algorithmic trading.2 This means that responsible executives should be held accountable for the use of AI which has been approved by the company; that there should be a requirement for the disclosure of the company's use of AI in its reports; that genuine human supervision should be part of the Board's processes. In the face of the emergence of AI, corporate law should focus on ensuring that its human representatives will continue overseeing business activities.</p>
    </sec>
  </body>
</article>
