Can Indian company law accommodate AI directors?

In 2014, a Hong Kong venture fund gave a machine-learning program named VITAL a seat on its board. Its managing partner said that the firm would not make an investment decision without VITAL’s sign-off. A decade later, Abu Dhabi’s International Holding Company inducted “Aiden Insight” as a non-voting AI board observer.

Indian boards are moving the same way. AI already shapes ESG analysis, risk mitigation and compliance monitoring at board level. The Securities and Exchange Board of India (SEBI) now holds regulated entities “solely responsible” for the output of any AI/ML tool they deploy, in-house or third-party, under the new Regulation 16C.

The question this piece asks is narrower than whether AI will influence Indian boardrooms; it already does. It is whether the Companies Act, 2013 could recognise an AI system as a director: a person who can be appointed to the board, owes fiduciary duties and can be held civilly and criminally liable for breaching them.

Section 149(1) requires the board to consist “only of individuals”. This is not incidental drafting. Section 149(3) requires at least one director to be a resident of India for 182 days - a requirement that only makes sense for a natural person. Section 152 requires a Director Identification Number obtained through Aadhaar-linked e-KYC. None of this can be satisfied by software.

More importantly, the duties in Section 166 presuppose a mind. “Good faith” is a state of actual honesty; “independent judgment” presupposes an agent whose deliberation is genuinely its own, not merely an output of its training data and its deployer’s objectives. An AI system cannot “avoid” a conflict of interest in the sense Section 166(4) contemplates, because it has no interest of its own to prefer. Any bias in its output is one embedded, deliberately or not, by whoever built or trained it.

The criminal liability provisions raise a sharper problem. Indian corporate criminal liability works by locating the mens rea of the natural persons forming a company’s “directing mind and will” and attributing it to the company. This is the basis of the alter-ego doctrine the Supreme Court applied in Iridium India Telecom Ltd v Motorola Inc (2011). An AI system supplies no such mind. Section 2(60)’s “officer who is in default” framework - which can mean imprisonment under provisions like Section 447 - is unworkable against an entity incapable of forming intent. Making an AI system a “director” would not create accountability; it would create a vacuum, since the officer nominally responsible would be structurally incapable of the mental state the Act's liability framework requires.

The same difficulty defeats the business judgment rule. In Miheer H Mafatlal v Mafatlal Industries Ltd (1997), the Supreme Court asked whether a scheme was one a reasonable businessman would take in the company’s best interest - an inquiry into motive and belief that only a natural person can sensibly undergo. An AI model has no belief to characterise as honest or self-interested, so there is nothing for a court applying the rule to interrogate.

If a board becomes “accustomed to act” on an AI system’s output the way DKV’s board did on VITAL’s, Section 2(60) already extends officer-in-default liability to “any person in accordance with whose advice, directions or instructions” the board acts. That liability lands on whoever configured, deployed or was accustomed to defer to the system — not on the system itself, since only a person can give “advice, directions or instructions” within the meaning of the Section. This shows Indian law already knows how to fasten accountability onto the humans standing behind an AI system, without any fiction of AI legal personhood.

Indian courts have been unusually willing to extend juristic personality beyond human beings - to deities and, briefly, to the Ganga and Yamuna, stayed by the Supreme Court within months. In Shiromani Gurdwara Prabandhak Committee v Som Nath Dass (2000), the Court traced this flexibility back to “convenience, justice and social utility”. But every entity in this line of cases exercises its personality through a natural person - a deity through its shebait, a company through its directors. The doctrine has never been a device for creating an autonomous locus of responsibility; it extends legal capacity to entities that cannot themselves exercise judgment by routing that judgment through an identified human who bears the resulting accountability. That is exactly the shadow-director structure above, not a case for AI directorship.

Comparative practice points the same way. The UK tightened, not loosened, its stance in 2015 by requiring every director to be a natural person. The EU’s AI Act requires high-risk AI systems to remain overseeable by natural persons rather than making them accountable in their own right. The European Parliament’s 2017 proposal for “electronic personhood” was abandoned precisely because it risked shielding the humans actually responsible for an AI system’s deployment.

Rather than inventing a directorship AI cannot hold, the Ministry of Corporate Affairs (MCA) could formally recognise a non-voting “AI board observer” status - mirroring what IHC already does informally - coupled with a named human director accountable for its mandate. Section 166 could be read to include a duty of algorithmic oversight: a director relying on AI-generated analysis should be able to show a reasonable understanding of its limitations and failure modes, not merely ratify its output. SEBI’s Regulation 16C model - responsibility resting on the deploying entity - could similarly extend to listed-company boards through Listing Obligations and Disclosure Requirements (LODR).

The Companies Act’s fiduciary architecture has always worked by locating a natural person’s state of mind and holding it to account. AI is going to sit ever closer to Indian boardrooms; the answer is not to make it a director, but to make sure whoever listens to it remains one.

Jenil Jain is an Associate at Chambers of Advocate Aarti Nimbalkar, Bombay High Court.