Advertisement
Advertisement
Skip to content
Follow Us on
Advertisement
TOP STORIES
CA, CS, CMA

Company Secretary’s Role in AI Governance, DPDP Compliance and Corporate Records

Summary: Artificial intelligence is transforming corporate governance and compliance by assisting Company Secretaries in preparing Board papers, minutes, statutory filings, compliance trackers, legal opinions, regulatory submissions and legal research. However, its growing use raises concerns regarding accuracy, confidentiality, accountability and the integrity of corporate records. The Supreme Court’s 2026 decisions in Pooja Ramesh Singh v. Jammu and Kashmir Bank Ltd. and Vijay Ghanshyam Gadiya v. Union of India, which set aside orders relying on non-existent legal authorities, underscore the importance of independent verification of AI-generated content. This article argues that the Company Secretary’s most valuable contribution to AI and data governance lies not in technical expertise but in the custody and reliability of corporate records. It examines professional responsibilities under the Companies Act, 2013, SEBI regulations and the Digital Personal Data Protection Act, 2023, focusing on Board meeting recordings, AI-generated minutes, unpublished price-sensitive information (UPSI), structured digital databases and the balance between statutory record retention and personal data erasure obligations. It also examines the independence restrictions on secretarial auditors of listed entities and the distinction between advisory and audit responsibilities. With substantive DPDP compliance obligations scheduled to commence on 13 May 2027, the article proposes a practical AI governance framework comprising Board-approved policies, a designated accountable officer, an AI Use Register, confidentiality safeguards, vendor controls, retention protocols and mandatory human verification. It explains the implications for practising Company Secretaries, whole-time Company Secretaries, secretarial auditors and founders, emphasising that AI can enhance efficiency but cannot replace professional judgment, statutory responsibility or human accountability. AI may help prepare a corporate document, but a named person must verify it, and the record must establish who did so.

Advertisement

Introduction

When the Citation Did Not Exist

Picture a reply to a show-cause notice, ready late in the evening before it is due. It is fluent and well organised, and it cites three judgments in support. Two of them are real. The third reads convincingly but was never delivered by any court. The draft was prepared with the help of an AI tool, reviewed for language and tone, and signed. Nobody checked the citations against the original reports, and nobody can now say who was expected to.

In July 2026 the Supreme Court set aside orders of the National Company Law Tribunal and the National Company Law Appellate Tribunal in an insolvency matter, because the orders relied on authorities that either did not exist or did not contain the passages attributed to them.[1] In September 2026 it set aside a customs penalty of about ₹425 crore for the same reason and sent the matter back for fresh adjudication by a different officer.[2]

The Court did not reject technology. In the customs case, it accepted that AI may be used “as an assistive tool to speed up the decision-making process”, but held that “assistance can never be substituted for adjudication”. Its image was memorable: “AI may well serve as training wheels but entrusting it with the pilot’s seat would be both imprudent and dangerous.” (Vijay Ghanshyam Gadiya, paras 6 to 8)

Both decisions concern adjudicators and advocates, not Company Secretaries. Their principle, however, applies by analogy to every document a company relies upon. Minutes, Board notes, compliance certificates, filings and replies to regulators are all read and acted upon by people who will not check them again. The analogy is closest where a Company Secretary represents a party before the Tribunal,[3] but it holds wherever the profession signs, certifies or keeps a record.

This article develops that argument through five questions:

  1. What law applies to AI and personal data in companies today?
  2. Where does the Company Secretary’s mandate come from, and where does it stop?
  3. How should three core records be governed when AI helps to produce them?
  4. What minimum framework can companies adopt before 13 May 2027?
  5. What does this mean in practice for the practising Company Secretary, the whole-time Company Secretary and the founder?

What Applies Today?

India has no statute on artificial intelligence. The India AI Governance Guidelines issued by the Ministry of Electronics and Information Technology in November 2025 are voluntary and rely on existing laws and sectoral regulators.[4] AI duties therefore reach companies through general law: the Companies Act, 2013, SEBI regulations, the Information Technology Act, 2000 with the CERT-In Directions, and the Digital Personal Data Protection Act, 2023 (DPDP Act).

Table 1: What binds, and from when[5]

Status Instruments
In force now Companies Act, 2013 and Secretarial Standards; SEBI (Prohibition of Insider Trading) Regulations, 2015; SEBI (Listing Obligations and Disclosure Requirements) Regulations, 2015; CERT-In Directions of 28 April 2022; section 43A of the IT Act and the SPDI Rules; definitions and institutional provisions of the DPDP Act
From 13 November 2026 Registration of Consent Managers under the DPDP Act and Rules (registration portal opened in November 2026)
From 13 May 2027 Substantive duties of Data Fiduciaries, rights of Data Principals, breach intimation and penalties under the DPDP Act; section 43A of the IT Act ceases to apply
Not yet binding MeitY AI Governance Guidelines (voluntary); SEBI guidelines on AI/ML (consultation, final guidelines awaited); RBI FREE-AI Report and draft Model Risk Management guidance

Three provisions of the DPDP framework bear directly on AI. First, a Data Fiduciary remains responsible for processing undertaken on its behalf by a Data Processor (section 8(1)), so outsourcing a task to an AI vendor does not outsource accountability. Second, where personal data is likely to be used to make a decision that affects the individual, the Data Fiduciary must ensure its completeness, accuracy and consistency (section 8(3)). Third, a Significant Data Fiduciary must observe due diligence to verify that algorithmic software it deploys is not likely to pose a risk to the rights of Data Principals (Rule 13(3)).

Sectoral regulators are moving in the same direction. SEBI has consulted on guidelines for the responsible use of AI and machine learning by market participants, and its Chairman stated in November 2026 that forthcoming guidelines will address disclosure of AI use, accountability of regulated entities and human oversight of critical decisions. The Reserve Bank of India has published the FREE-AI Report and a draft guidance on model risk management that would require Board-approved frameworks and full accountability for vendor models.[6] Until finalised, these indicate the direction of travel; they are not binding.

Where Does the Company Secretary’s Mandate Come From?

The DPDP Act does not mention the Company Secretary. Its duties fall on the Data Fiduciary, and its penalties on the person in breach (section 33). Unlike the Companies Act, it contains no officer-in-default provision.

The mandate comes instead from company law. The Company Secretary reports to the Board on compliance with the Act, the rules and other laws applicable to the company, and the functions prescribed under the Rules include guiding directors on their duties, assisting the Board in good corporate governance and best practices, and discharging other duties assigned by the Board.[7] Section 205(2) preserves the duties of the Board and of others under the law. In listed entities, the compliance officer’s responsibility for conformity with regulatory provisions under Regulation 6 of the LODR Regulations adds a further anchor.

Two practical consequences follow. First, the Board should define by resolution what the Company Secretary is expected to do in AI and data governance. Otherwise the Company Secretary may carry responsibility without authority. Second, the question of the Data Protection Officer must be considered separately. Only a Significant Data Fiduciary must appoint one, and that officer must be based in India, be responsible to the Board and act as the point of contact for grievances (section 10(2)(a)). Nothing in the Act prevents a Company Secretary from holding the role. It is submitted, however, that where both roles are held by one person, the Board should record the reporting lines, because the Company Secretary may later be the person advising the Board on whether the Data Protection Officer’s function is working.

For directors, AI changes nothing in principle. The duty to exercise due and reasonable care, skill and diligence and independent judgment under section 166(3) is not discharged by relying on a summary, a risk score or a draft that no one has checked.

First Record: The Boardroom

Who is in the Meeting?

AI notetakers are now routinely offered by video-conferencing platforms and by independent vendors. For meetings held through video conferencing or other audio-visual means, Rule 3 of the Companies (Meetings of Board and its Powers) Rules, 2014 requires the Chairperson and the Company Secretary to take due and reasonable care to ensure that “no person other than the concerned director” attends or has access to the proceedings. Each director confirms this at the roll call, and the names of other persons present are recorded.[8]

The concern is not the platform that hosts the meeting. It is a separate bot that joins the call, a director’s personal transcription app, or a transcript stored on a vendor’s servers outside the company’s control. It is submitted that these give a third party access to the proceedings in substance, even if no human attends. A workable practice is to permit only a tool approved by the Board, operating within the company’s own environment, and to announce its use at the roll call so that it is recorded. For meetings held in person, the same result follows from the confidentiality of Board proceedings and the provisions of section 118.

Is the Transcript a Recording?

For video-conferencing meetings, the recording must be stored and marked as part of the company’s records “at least before the time of completion of audit of that particular year”.[9] An AI transcript may or may not be such a recording. The Rules do not say. A company should decide the question in its policy and fix how long transcripts are kept, rather than leaving the answer to the default settings of a vendor’s product.

Which Record Prevails?

Minutes must contain a fair and correct summary of the proceedings and are evidence of them.[10] The Chairman has absolute discretion to exclude matter that is defamatory, irrelevant or immaterial, or detrimental to the interests of the company. A recording kept alongside minutes is contemplated by the Rules and is not improper. But a verbatim AI transcript that is widely circulated or kept indefinitely may, in a dispute, undercut the practical effect of that discretion and may be called for in litigation. It is submitted that this is the strongest reason for a clear policy on transcripts.

Who Verified the Minutes?

An AI-generated summary may be a useful first draft. It cannot decide what a director meant. The Company Secretary should personally verify the points that carry legal weight: the names of directors dissenting from or not concurring with a resolution, disclosures of interest, recusals and the exact text of resolutions. Changes after the minutes are entered and signed attract the risk of section 118(12), so the verification must happen before signature, not after.

Minutes also contain personal data. A vendor providing transcription will ordinarily be a Data Processor and must be bound by a contract that includes reasonable security safeguards. A vendor that uses inputs for its own purposes, such as training its models, may itself be a Data Fiduciary. The contract should therefore prohibit such use.

Second Record: Unpublished Price Sensitive Information

Is Every Use of AI a Leak?

It is sometimes said that any unpublished price sensitive information (UPSI) entered into an AI tool is a violation. The Regulations are more precise. Regulation 3(1) prohibits an insider from communicating, providing or allowing access to UPSI, except in furtherance of legitimate purposes, performance of duties or discharge of legal obligations. The Board must have a policy determining legitimate purposes, a recipient for a legitimate purpose is treated as an insider and must be given notice to maintain confidentiality, and the structured digital database must record the sharing.[11]

It is submitted that the following questions decide the matter on the facts:

  1. Can the vendor’s staff or systems see, retain or train on the information?
  2. Does the Board’s policy on legitimate purposes cover the use?
  3. Has the vendor been given notice to maintain confidentiality?
  4. Is the sharing entered in the structured digital database?

On this test, a public consumer AI tool should be treated as prohibited for UPSI: there is no notice, no confidentiality undertaking and no control over retention. An enterprise tool operating under contractual and technical limits may be defensible, if the policy and the database are brought up to date. No SEBI position on AI tools has been traced, so companies should adopt the cautious view.

The Structured Digital Database

The database “shall not be outsourced” and must be maintained internally with time stamping and audit trails. SEBI’s FAQs do not classify cloud storage as internal or outsourced. They make the Board solely accountable for data on the cloud and treat software hosted on a vendor’s server as contrary to the requirement where the vendor may have access.[12] An AI assistant connected to the database raises the same concern and should not be permitted without the same scrutiny. Regulation 9A, which requires internal controls to prevent insider trading and an inquiry procedure for leaks, is the natural place to record these controls.

When a Prompt Becomes a Breach

The DPDP Act adds a second risk. A personal data breach includes unauthorised processing and accidental disclosure that compromises the confidentiality, integrity or availability of personal data. From 13 May 2027, pasting employee or customer data into an unapproved tool may therefore require intimation to the Data Protection Board and to affected individuals, with a detailed report within 72 hours. Failure to take reasonable security safeguards can attract a penalty of up to ₹250 crore, and failure to intimate a breach up to ₹200 crore.[13] Listed entities must also report cyber security incidents in the quarterly corporate governance compliance report under Regulation 27(2) of the LODR Regulations, and consider whether Regulation 30 requires disclosure.

Third Record: Keep or Erase?

Data protection is often understood as an instruction to delete. The Company Secretary, by contrast, is the custodian of registers and minutes that the law requires to be kept. The DPDP Act resolves this conflict itself. The duty to erase personal data yields where retention is necessary for compliance with any law in force, and the right to erasure is subject to the same exception. The Rules add a minimum retention period of one year for processing logs and associated data.[14]

The real gap lies elsewhere: in the AI residue that has no statutory home. Transcripts, summaries, prompt histories and search indexes built over minute books have no prescribed retention period. A policy should set one, require processors to erase on the same timetable, and suspend deletion when a dispute, inquiry or investigation is foreseeable.

Table 2: Keep or erase

Record Retention basis AI governance point
Minutes of Board meetings Preserved permanently (Secretarial Standard on Meetings of the Board of Directors) Keep in the minute book, not in a tool’s workspace
Recording of a video-conferencing meeting At least until completion of that year’s audit (Rule 3(2)(d)) Decide by policy whether an AI transcript is covered
Structured digital database At least eight years, or longer during proceedings (PIT Regulation 3(6)) Not outsourced; restrict any AI access
Processing logs and associated data Minimum one year (DPDP Rules 6(1)(e), 8(3)) Vendor contract must cover retention and erasure
AI transcripts, summaries and prompts None prescribed Fix by policy; hold during disputes and inquiries

The Limits of the Role: Reporting without Auditing Your Own Work

Within the company, section 205(1)(a) supports a periodic report by the Company Secretary to the Board on AI and data compliance. The Directors’ Responsibility Statement under section 134(5)(f) requires directors to confirm that proper systems exist to ensure compliance with all applicable laws, and from FY 2027-28 those laws will include the substantive provisions of the DPDP Act.

For the secretarial auditor, two questions arise. The first is wording. Paragraph (vi) of Form MR-3 refers to other laws applicable specifically to the company, and ICSI’s Guidance Note explains these by reference to sector-specific laws.[15] It is submitted that for most companies the DPDP Act is better addressed through the confirmation on adequacy of systems and processes, while a company whose business is built on personal data may warrant naming it under paragraph (vi). Since FY 2026-27 ends before 13 May 2027, FY 2027-28 will be the first audit year in which the substantive duties apply.

The second question is independence, and it is the one most likely to be overlooked. Under Regulation 24A(1B) of the LODR Regulations, the secretarial auditor of a listed entity shall not render, to the entity, its holding company or its subsidiary, services that include the “design and implementation of any compliance management system, information system, policy framework, systems or processes for compliance”.[16] The restriction applies to services rendered directly or indirectly, including through partners, associate entities and entities using the same name or brand. ICSI’s list of permitted and prohibited services also shows acting as an advisor for legal or procedural matters as not permitted.[17]

The consequence is clear. A practising Company Secretary who designs or advises on the AI and data governance framework of a listed entity should not also be its secretarial auditor, and the firm’s associates should be treated the same way. For unlisted companies, the general principles of independence and professional conduct apply.

From Practice: Lessons from Daily Use

The proposals in this article are not theoretical. In the author’s practice, AI tools are used every day: to prepare first drafts of Board notes, resolutions, replies to notices and opinions; to research regulatory changes and summarise circulars, judgments and lengthy notes; to build checklists, compliance trackers and calendars; and to transcribe and summarise meetings. The gains in speed are real, particularly in research and in reducing long documents to their essentials.

Three lessons have emerged. First, citations cannot be trusted until checked. Tools have produced sections, circulars and judgments that were wrong or did not exist, in the same confident language as those that were correct. Second, a fluent answer is not a correct one. Drafts have read well while missing a fact, an exception or a proviso that changed the conclusion. Third, the greatest risk is human rather than technical: the instinct of a busy team member is to paste a client’s document into whichever tool is at hand.

The practice has responded with five rules:

  1. Every AI-assisted draft is reviewed by a qualified Company Secretary before it leaves the office.
  2. Every section, circular and judgment is verified against the bare Act or the original source.
  3. Client names and confidential information are not entered into public AI tools.
  4. AI is not used to form audit conclusions or to issue certificates.
  5. No AI output is relied or acted upon as an opinion until it has been verified.

None of these rules requires technology; each requires discipline. They are also a matter of professional conduct, since client confidentiality and due diligence are duties of the profession. In substance, they are the verification rule proposed below, applied to the profession’s own work.

A Workable Minimum before 13 May 2027

Most companies do not need an elaborate AI governance programme. They need four things: one owner, one Board note, one register and one rule. The owner is a named executive, often the Company Secretary, who coordinates the work. The Board note sets out the company’s current use of AI and the gaps. The register records every AI tool in use, and the rule is simple: no filing, Board paper, reply to a regulator or certificate relies on AI-generated text unless a named person has checked it against the original source.

Table 3: The AI Use Register

Register field What it captures
Tool and owner Name of the tool, the business function using it and the accountable person
Purpose and data What it is used for; whether personal data or UPSI is involved
Vendor and location Contract terms on access, training, retention and where data is held
Retention How long inputs, outputs and transcripts are kept, and how they are erased
Human reviewer and approval Who verifies the output; date of approval by the Board or committee

Table 4: A readiness timeline

Period Priority actions
By December 2026 Name the owner, place a note before the Board, prepare the inventory of tools, and bar UPSI and personal data from public AI tools
January to February 2027 Approve the AI policy; update the legitimate-purpose policy and the database process; adopt a protocol for minutes and transcripts; review vendor contracts
March to April 2027 Approve the retention matrix; rehearse breach response across the DPDP Rules, CERT-In and the LODR Regulations; issue notices for existing consents
By 13 May 2027 Controls in operation; first report to the Board

The framework should be scaled to the company. A listed entity should add the changes to its insider trading policy and database. A non-banking financial company should follow the RBI’s guidance as it is finalised. A smaller company may need no more than a one-page policy, the register and a check of its vendors’ terms.

What This Means in Practice: Four Roles

The same principle applies differently depending on where the professional sits.

Table 5: How the framework helps each role

Role Where it helps What to watch
PCS as secretarial auditor A clear audit trail to test: AI Use Register, Board note, minutes protocol, verification sign-offs Do not design or advise on the framework of a listed entity you audit (Regulation 24A(1B))
PCS as Boardroom advisor A structured service: AI policy, notetaker protocol, legitimate-purpose policy, vendor terms, breach drill Keep advice and audit separate across the firm and its associates
Whole-time Company Secretary Authority from a Board resolution; one register and one verification rule to run Do not accept responsibility without a defined scope
Founder Investor-ready records and fewer surprises in due diligence Public AI tools and unchecked drafts in filings, Board papers and data rooms

The Practising Company Secretary in Audit

Secretarial audit depends on evidence. Where AI is used in preparing minutes, registers or filings, the auditor needs more than the final document: the auditor needs to know which tool was used, who verified the output and when. The register and the verification rule proposed above give the auditor that trail. From FY 2027-28, the auditor can test them in the same way as any other compliance system and report on them through the confirmation on adequacy of systems and processes in Form MR-3. Where the trail is missing, the absence is itself an observation worth recording.

The Practising Company Secretary in Boardroom Advisory

For the practising professional who is not the auditor, AI and data governance is a natural advisory service, because it builds on work the profession already does: drafting policies, settling Board processes, reviewing contracts and preparing compliance calendars. A practical engagement can cover the AI policy, the protocol for notetakers and transcripts, the update to the legitimate-purpose policy and the database process, the review of vendor terms and a breach-response drill. Two limits apply. In a listed entity, this work should not be done by the secretarial auditor or by anyone connected with the auditor, for the reasons explained earlier. And the advisor should not overstate the law: where a position is a matter of interpretation, the Board should be told so.

The Whole-time Company Secretary

For the in-house Company Secretary, the most important step is the first one: a Board resolution that sets out the scope of the role in AI and data governance. With that authority, the Company Secretary can maintain the register, apply the verification rule to every Board paper and filing, keep the minutes protocol, and report to the Board under section 205(1)(a). In a listed entity, the same work supports the compliance officer’s responsibilities and strengthens the Board’s basis for the Directors’ Responsibility Statement. Without a defined scope, the Company Secretary risks being held responsible in practice for decisions that others take.

The Founder

Founders are often the earliest users of AI tools, and the least likely to have a policy on them. For a founder, the framework is not an additional burden but a way to avoid three common problems. The first is a draft filing, investor update or Board paper that contains an error nobody checked. The second is confidential or personal data entered into a public tool by a team member. The third is a due diligence exercise that reveals records which cannot be traced to their source. A one-page policy, the register and a single named reviewer cost little, and they make the company’s records more reliable when an investor, a lender or a regulator asks for them.

Conclusion: The Question That Remains

AI will continue to draft more of what companies say and record. The profession does not need to resist that, nor does it need to become a profession of technologists. The Company Secretary’s responsibility is older and simpler: to ensure that the corporate record is reliable, and that it shows who decided and who checked.

For every AI-assisted document, one question should be asked before it is signed, filed or relied upon: who verified this? If the answer is a named person who checked it against the source, the technology has served the company well. If there is no answer, the Supreme Court has already shown what follows. Guidance from the Institute on the use of AI in secretarial records would help members apply that question consistently.

Notes:

[1]Pooja Ramesh Singh v. Jammu and Kashmir Bank Ltd., 2026 INSC 668 (Supreme Court, 2 July 2026).

[2]Vijay Ghanshyam Gadiya v. Union of India, 2026 INSC 947 (Supreme Court, 2 September 2026), paras 6 to 8.

[3]Section 432, Companies Act, 2013.

[4]Ministry of Electronics and Information Technology, India AI Governance Guidelines, 5 November 2025.

[5]Notification G.S.R. 843(E) dated 13 November 2025 and the Digital Personal Data Protection Rules, 2025, notified vide G.S.R. 846(E).

[6]SEBI, Consultation Paper on Guidelines for Responsible Usage of AI/ML in Indian Securities Markets, 20 June 2025; remarks of the SEBI Chairman reported by PTI on 15 November 2026; RBI, Report of the Committee on the Framework for Responsible and Ethical Enablement of Artificial Intelligence (FREE-AI), 13 August 2025; RBI, draft Guidance on Regulatory Principles for Model Risk Management, 2026.

[7]Section 205(1)(a), Companies Act, 2013, and Rule 10 of the Companies (Appointment and Remuneration of Managerial Personnel) Rules, 2014.

[8]Rule 3(2)(e), 3(4)(d) and 3(5)(a), Companies (Meetings of Board and its Powers) Rules, 2014.

[9]Rule 3(2)(d), Companies (Meetings of Board and its Powers) Rules, 2014.

[10]Section 118(2), (5), (6) and (7), Companies Act, 2013.

[11]Regulation 3(1), (2A), (2B) and (5), SEBI (Prohibition of Insider Trading) Regulations, 2015.

[12]Regulation 3(5), SEBI (Prohibition of Insider Trading) Regulations, 2015; SEBI, Frequently Asked Questions on the SEBI (Prohibition of Insider Trading) Regulations, 2015, FAQs 7 and 8 (version of 31 December 2024).

[13]Section 2(u), section 8(6) and section 33 read with the Schedule, Digital Personal Data Protection Act, 2023; Rule 7, Digital Personal Data Protection Rules, 2025.

[14]Section 8(7) and its illustration, and section 12(3), Digital Personal Data Protection Act, 2023; Rules 6(1)(e) and 8(3), Digital Personal Data Protection Rules, 2025.

[15]ICSI, Guidance Note on Secretarial Audit; Form No. MR-3, Companies (Appointment and Remuneration of Managerial Personnel) Rules, 2014.

[16]Regulation 24A(1B), SEBI (Listing Obligations and Disclosure Requirements) Regulations, 2015, read with Annexure 3 to SEBI Circular No. SEBI/HO/CFD/CFD-PoD-2/CIR/P/2024/185 dated 31 December 2024, as carried into the LODR Master Circular.

[17]ICSI, list of services which a Secretarial Auditor can or cannot render, 24 April 2025.

Advertisement

Author Info

Dr. CS RONAK JHUTHAWAT
Qualification: CS
Company: Ronak Jhuthawat &; Co.
Location: Udaipur, Rajasthan
Articles Published: 55

Join TaxGuru's Network for the latest updates on Income Tax, GST, Company Law, Corporate Laws and other related subjects.

Leave a Reply

Your email address will not be published. Required fields are marked *