As of November 2026, an Indian business using AI sits under five live rules and two coming deadlines. Live: the Digital Personal Data Protection Act 2023 with its Rules notified on 13 November 2025 and phasing in over 18 months; the IT Rules amendment in force since 20 February 2026 requiring AI-generated audio, image and video to be prominently labelled; the India AI Governance Guidelines of November 2025, which create no licence but set the documentation regulators expect; sector rules from RBI (the FREE-AI framework, August 2025) and SEBI for financial firms; and the consumer authority’s 2024 guidelines on coaching advertisements. Coming: the DPDP consent-manager framework from 12 November 2026 and the full DPDP obligations by around May 2027. For exporters, the EU AI Act’s high-risk duties were deferred to December 2027 by the Digital Omnibus, while its prohibitions and transparency duties stand. For a 20-person business the practical list is short: a notice and consent for every WhatsApp list, a deletion path, a label on AI-generated media, vendor data terms you have read, and a one-page data-flow diagram.
Every rule that touches a small or mid-sized Indian business using AI, dated, with a link to the source and one line on what it means for a 20-person company. Newest first, upcoming deadlines flagged, and where the status is "reported, not yet in the Gazette" the entry says so. Maintained because the rules moved faster in 2025–26 than any vendor page kept up with.
Each entry has a date, what happened, a link to the primary source where one is public, and a line on what it means for a business of about twenty people: a clinic, a CA firm, a coaching institute, a brokerage. The "means for you" line is ours; the rest is the regulator’s. Where an entry’s status is reported by law firms or the press but we could not confirm the Gazette text, it says so, and you should check before relying on a date.
The page is dated at the top and refreshed monthly. If a rule you rely on is missing, the enquiry form at the foot of every page reaches us.
Upcoming deadlines are marked; everything else is in force or published as dated.
DPDP: full obligations in force (end of the 18-month phase-in)
Law-firm readings of the DPDP Rules 2025 put the end of the phased transition at about 18 months from notification, by when the detailed notice requirements, consent duties, data-principal rights (access, correction, erasure), breach notification and the significant-data-fiduciary duties all apply. The exact end date depends on the Rules’ schedule; verify against the notification.
Means for you: By then every WhatsApp list, booking record and enquiry form needs a plain-language notice, a consent you can show, and a way for a person to ask for deletion and get it. Start the notice and the deletion path now; they take a quarter to get right.
SourceDPDP: consent-manager registration framework becomes operational
One year after notification, the framework for registering consent managers (the intermediaries through which individuals can manage consents across businesses) switches on. The Data Protection Board itself was constituted at notification.
Means for you: Nothing to file for a small business. It is the signal that the second phase is near: if your consent records are still "they messaged us first", fix that this quarter.
SourceEU AI Act: Digital Omnibus on AI approved, high-risk duties deferred
Reported Council approval of the Digital Omnibus on AI moves the standalone high-risk (Annex III) obligations from 2 August 2026 to 2 December 2027 and product-embedded (Annex I) obligations to 2 August 2028. Prohibitions (in force since February 2025) and general-purpose model duties (August 2025) are unchanged. Sources conflict on whether Article 50 transparency duties (chatbot disclosure, synthetic-content marking) still apply from August 2026 or get a short grace period; Official Journal publication was not confirmed at the time of writing.
Means for you: If you sell software with AI features into the EU, the heavy conformity work moved out to late 2027, but telling users they are talking to an AI and marking generated media is the safe assumption from August 2026. If you only serve Indian customers, this does not apply to you.
SourceMeitY: draft further amendment on synthetic-content labels
A draft proposing that the label on synthetically generated information be continuously and clearly visible for the whole duration of the content, with comments invited to 7 May 2026. Whether it has been notified was not confirmed at the time of writing.
Means for you: If it lands, an AI-generated reel or voice-over in your ads carries a persistent on-screen label, not a caption at the end. Design marketing creative on that assumption.
SourceIT Rules Amendment 2026 in force: label AI-generated content, faster takedowns
Notified 10 February 2026 (G.S.R. 120(E)) and in force ten days later. Defines "synthetically generated information" (audio, image or video that appears real), requires it to be prominently labelled with persistent metadata, shortens takedown timelines for intermediaries, and confirms that removing such content under the rules does not cost an intermediary its safe harbour.
Means for you: The duties fall on platforms, but the label falls on what you upload: an AI-generated product image, testimonial video or voice-over in your marketing should be labelled, and platforms will increasingly ask you to declare it. An agent’s text replies are not audio-visual and are not covered by this definition.
SourceDPDP Rules 2025 notified; 18-month phased rollout begins
The Rules under the Digital Personal Data Protection Act 2023 were notified on 13 November 2025 (Gazette 14 November). Institutional provisions, including the Data Protection Board, took effect immediately; consent-manager registration follows at twelve months and the substantive obligations by about eighteen.
Means for you: The clock started. The Act’s penalties run to ₹250 crore for the worst breaches, and the Rules now say what notice, consent and deletion have to look like. A twenty-person business needs the notice, the consent record and the deletion path, and a named person who answers requests.
SourceIndia AI Governance Guidelines released
MeitY’s guidelines take a techno-legal approach: no new AI law or licence, reliance on existing statutes (IT Act, DPDP, consumer and sectoral law), principles for responsible use, and new institutions (an AI governance group, a safety institute, an incident database). Voluntary for business, binding in tone for regulators.
Means for you: No permission to get. What the guidelines ask for is what a well-run pilot produces anyway: a record of what the system does, who is accountable, what data it sees, and how a person can appeal. Keep the data-flow diagram and the logs; they are your compliance file.
SourceRBI: FREE-AI committee report on AI in the financial sector
The committee’s Framework for Responsible and Ethical Enablement of AI sets out 26 recommendations for regulated entities: board-approved AI policies, risk-tiering, incident reporting, consumer protection, and a tolerant posture towards first-time errors made in good faith. Recommendations, with RBI action to follow.
Means for you: If you are an NBFC, a fintech or a vendor to a bank, expect to be asked for an AI policy, a model inventory and an incident process. If you simply bank with one, nothing changes.
SourceWhatsApp Business Platform: per-message pricing replaces per-conversation pricing
Meta moved the Business Platform to per-message pricing for template messages, with marketing, utility and authentication categories priced differently by country, and utility messages free inside an open customer-service window.
Means for you: Your reminder and recall costs are now per message sent, and marketing templates cost more than utility ones. Write reminders as utility templates, keep marketing for people who opted in, and read the provider’s bill against Meta’s rate card.
SourceSEBI: consultation on responsible use of AI and machine learning in securities markets
A consultation paper on guidelines for AI/ML use by regulated entities, following earlier amendments making intermediaries responsible for the outputs of any AI tools they use with investors. Final guidelines to follow consultation.
Means for you: A broker, adviser or distributor owns what its AI tells a client. If an agent answers investor questions for you, it answers from approved material only and the firm keeps the log.
SourceTRAI: tighter rules on commercial SMS and calls
Amendments to the Telecom Commercial Communications Customer Preference Regulations: stricter consent and header rules for promotional SMS and calls, a longer complaint window, and penalties for telemarketers and senders.
Means for you: Promotional SMS and calls need registered headers and templates on the DLT system and a consent you can show. WhatsApp is outside TRAI’s scope, but Meta’s own opt-in and template rules do the same job there.
SourceCCPA: guidelines against misleading advertisements in the coaching sector
The Central Consumer Protection Authority’s guidelines bar coaching centres from claims about selection rates, ranks and guaranteed results without verifiable disclosure, and treat testimonials and success stories as advertisements.
Means for you: Every WhatsApp broadcast and every landing page a coaching institute’s CRM points at is an advertisement under these rules. An agent that answers enquiries must not claim results; it quotes published fees and timings and leaves the rest to a counsellor.
SourceEU AI Act enters into force
Prohibited practices applied from 2 February 2025 and general-purpose model duties from 2 August 2025; high-risk obligations were scheduled for August 2026 and have since been deferred (see the June 2026 entry).
Means for you: Relevant only if you sell into the EU. If you do, the prohibitions already apply and the transparency duties are the near-term work.
SourceMeitY revised AI advisory: label, disclose, no permission regime
A revised advisory replaced the 1 March version that had asked for government permission before deploying under-tested models. The revision dropped the permission requirement and kept the asks: label under-tested or unreliable AI, label deepfakes and synthetic media, and tell users about the limits of what they are using.
Means for you: The "label and disclose" posture every later rule builds on started here. An agent that says it is an AI, says what it cannot do, and hands over to a person is already doing what the advisory asked.
SourceDigital Personal Data Protection Act 2023 enacted
India’s first comprehensive personal-data law: consent as the default basis for processing, obligations on data fiduciaries, rights for data principals, a Data Protection Board, and penalties up to ₹250 crore per instance. Operational only once the Rules were notified in November 2025.
Means for you: The base law your customer, patient, student and client data sits under. Everything above about notices, consent and deletion comes from here.
SourceOne: a notice and a consent for every list. Everyone you message on WhatsApp, SMS or email should have been told, in plain words, what you will send and how to stop it, and you should be able to show when they agreed. The Business Platform’s opt-in rules and the DPDP Rules ask for the same thing.
Two: a deletion path. When someone asks you to delete their data, someone in the business knows how, and does it across the record system, the WhatsApp provider and the sheet. Write the three steps down.
Three: a label on AI-generated media. Any image, video or voice you generated and publish gets a visible label. Text your agent writes does not fall under the February 2026 definition, but saying "this is an automated assistant" at the start of a conversation is the posture every rule since March 2024 has asked for.
Four: vendor data terms you have read. For every AI tool that sees customer data, know whether it trains on your data, where it stores it and for how long. Business and API tiers usually say no to training; consumer tiers often do not.
Five: a one-page data-flow diagram and a log. What each system and agent reads, where it goes, who can see it, how long it stays. It is what the AI Governance Guidelines describe, what the DPDP Rules will expect you to be able to show, and what a customer or an auditor asks for first.
Where the rules themselves live. Law-firm summaries are useful; these are what they summarise.
The Act, the Rules 2025 and the phase-in schedule. The one page to bookmark.
The November 2025 guidelines and the institutions they set up.
The FREE-AI committee report and any circulars that follow it.
The AI/ML consultation and the intermediaries regulations that make firms responsible for their tools.
The coaching-sector advertising guidelines and the general misleading-advertisement rules.
The TCCCPR amendments governing promotional SMS and calls.
The Act, its timeline and the Digital Omnibus changes, for anyone selling into the EU.
The per-message rate card and the category rules every provider passes through.
Cite this page: SigmaticAI, "AI regulation tracker for Indian businesses, 2026", appsigmaticai.com/library/ai-regulation-tracker-india, updated 1 November 2026.
Link to it rather than copying the table; the numbers are refreshed here and the sources are one click away.
The AI Agent Pilot puts one agent on your data in 2–3 weeks for $4,500 (₹3,50,000) fixed, measured against your own baseline. Thirty minutes to scope it, free.