Part of our guide: Is AI Calling Legal in India? TRAI, DND, DPDP and Consent (2026)
Key takeaways
- TRAI now defines A2P calls as calls started by software or an automated platform without direct human dialling, including artificial-voice calls.
- Businesses must declare A2P use and their calling numbers to their telecom provider in advance; undeclared A2P calls count as spam.
- A termination charge of up to 5 paise per minute applies to A2P calls, with designated commercial number series exempt.
- Telecom operators must use AI/ML to flag likely spam numbers and share them; flagged senders face graded action up to one-year disconnection.
- Reported start dates are phased: 30, 60 and 90 days after Gazette publication.
On 18 September 2026, TRAI announced the Telecom Commercial Communications Customer Preference (Third Amendment) Regulations, 2026. For anyone running automated or AI calls, it is the most important change in years: TRAI rules for AI calls now include a formal definition of automated (A2P) calls, an advance declaration requirement, and a per-minute termination charge on automated calls made outside the designated commercial number series. Here is what the official text and reputable reporting say, with dates and links.
What did TRAI change in September 2026?
TRAI's Third Amendment to the TCCCPR, announced on 18 September 2026, brings automated calls under a declaration regime, adds a termination charge on A2P calls, writes AI-based spam detection into the regulations, tightens consent and inquiry rules, and protects the 140 and 1600 number series from blanket spam-tagging by apps.
The changes below come from TRAI's press release of 18 September 2026. The draft was issued on 13 March 2026, with comments, counter-comments and an open house discussion on 3 June 2026 before the final version.
| Change | What it says | Why it matters for AI calling |
|---|---|---|
| A2P call definition | Voice calls initiated by an application, software system or automated platform without direct human dialling, including autodialling, robocalls and pre-recorded or artificial voice | Automatic AI calling fits this description |
| Prior declaration | Senders declare A2P use and the numbers (CLIs) they will use to their telecom provider | Undeclared A2P calls are treated as spam |
| Termination charge | Up to 0.05 per minute, levied by the receiving operator on the originating one; designated series and authorised calls exempt | A possible new per-minute cost |
| AI/ML detection | Operators identify numbers with a high probability of spam and share them | Detection no longer waits for complaints alone |
| Inquiry follow-up | Follow-up after a written or digital inquiry allowed for seven days; inquiry kept in verifiable form | Keep a record of how each lead reached you |
| Legacy consent | Valid only if obtained through verifiable means and registered on the operator's digital platform | Old "consent" without proof may not count |
| Number series | Apps may not blanket-block or spam-tag 140xx, 1600xx and 1601xx calls | Regulated series get protection from blanket tagging |
The press release does not name a currency for the charge; MediaNama's report on the March draft describes it as 5 paise per minute, and illustrates that one million one-minute automated calls a month could add up to ₹50,000.
Do TRAI rules for AI calls cover voice agents?
Yes, in substance. The 2026 definition of A2P calls covers calls started by software without direct human dialling and explicitly includes artificial-voice calls. An AI voice agent that dials leads from a sheet or CRM matches that description. The amendment, as reported, does not name "AI voice agents", so confirm your exact set-up with your telecom provider.
This isn't entirely new
Robocall rules in India predate 2026. The consolidated TCCCPR text required senders to notify their originating access provider in advance, in writing, about the use of auto-dialers or robo-calls and their objective. It defined robo-calls as calls using an artificial or prerecorded voice without a human on the calling side. You can read this in TRAI's consolidated TCCCPR document (May 2026). What the Third Amendment adds is a clear A2P definition, the consequence (undeclared calls count as spam) and the termination charge.
What declaration means in practice
According to IndiaLaw's analysis of the final rules, entities making automated or pre-recorded calls must declare their full series of calling numbers to their telecom providers, and these are recorded on the DLT (distributed ledger) platform. A commercial call from an undeclared series is treated as unsolicited commercial communication. If an AI calling provider places calls for you, ask in writing which numbers the calls use and who is responsible for the declaration.

When do TRAI's 2026 anti-spam rules take effect?
The TRAI press release says the regulations were introduced on 18 September 2026 but gives no commencement date. Legal and financial reporting describes a phased start counted from Gazette publication: general provisions and updated DND rules after 30 days, A2P declarations after 60 days, and appeals and multi-operator AI matching after 90 days.
That schedule is reported by IndiaLaw (6 October 2026) and Upstox (29 September 2026). Neither gives the Gazette publication date, and we have not seen it, so we are not stating exact calendar dates. The safe assumption: plan to have your automated calling declared well before the 60-day mark.
Upstox also reports that preferences under the old "fully blocked" DND category move to a new BLOCK PROMO option, which blocks promotional communication but not transactional, service or government calls.
What are 140 and 160 series numbers?
The 140 series carries regulated promotional calls. The 1600 series, often searched as "160 series numbers", carries service and transactional calls from regulated entities such as banks, insurers, mutual funds and government bodies, allotted by the Department of Telecommunications. Ordinary 10-digit numbers are restricted from telemarketing under TRAI's February 2025 amendment.
Key dates for the series:
| Date | What happened | Source |
|---|---|---|
| 12 Feb 2025 | Second Amendment: promotional calls on 140 series, 1600 series for service/transactional calls, normal 10-digit numbers restricted from telemarketing | TRAI press release 11/2025 |
| 1 Jan – 1 Mar 2026 | Migration deadlines for RBI-regulated banks and NBFCs to 1600 numbers for service and transactional calls, by entity type | Storyboard18 |
| 15 Feb 2026 | Deadline for IRDAI-regulated insurers (direction of 16 December 2025) | PTI via The Week |
| 18 Sep 2026 | Apps barred from blanket blocking or spam-tagging 140xx, 1600xx and 1601xx calls | TRAI press release, Sep 2026 |
If you are a bank, NBFC or insurer, your service calls belong on 1600 numbers. For other businesses, the practical rule is simpler: promotional calling should not go out from ordinary mobile numbers.
How will telecom operators catch spam callers now?
Operators must run AI/ML systems that flag numbers likely to be sending spam and share them across networks. A TRAI direction of 27 February 2026 required this within 30 days, and the Third Amendment writes it into the regulations. Flagged senders face graded action, from KYC re-verification to one-year disconnection.
According to Khaitan & Co's note on the February direction, systems watch call volume, velocity, diversity, duration and timing patterns. A flagged number is shared with the originating operator within two hours, and if five or more numbers linked to one sender are flagged within ten days, action follows. Under the Third Amendment, three or more unique complaints within ten days, combined with an AI flag, can also trigger action.
On penalties, reports differ slightly on the sequence. Upstox reports a 15-day bar on outgoing services for a first violation and disconnection of all telecom resources, including SIP trunks and SIMs, for one year for later violations. IndiaLaw describes KYC re-verification for a first violation, physical verification and a 15-day suspension for a second, and disconnection with a one-year blacklist for repeated violations. It also cites TRAI data that AI systems flagged about 771 crore calls as suspected spam in April 2026.
Those parameters are where the TRAI anti-spam rules 2026 touch everyday AI calling, and why TRAI rules for AI calls are not just about declarations. A sudden burst of thousands of very short calls from a few numbers looks like spam whether or not it is. Steady volume to people who asked to hear from you, with real conversations, looks different.
Watch your share of calls ending within a few seconds. A rising share of very short calls hurts conversions and is the kind of pattern AI spam detection looks at. Clean your list and fix your opening line before adding volume.
What should businesses running AI calls do now?
Declare automated calling and its numbers to your telecom provider, keep verifiable records of every inquiry and consent, call inquiry leads within the seven-day window, avoid ordinary 10-digit numbers for promotional calling, and pace volume so your traffic looks like conversations, not blasts.
- Map your calls. List which campaigns are promotional, which are service or transactional, and which are automated.
- Declare A2P use. Ask your telecom provider, or your AI calling provider, to confirm the declaration and the numbers covered, in writing.
- Keep the inquiry trail. Store the form, ad lead or message that brought each lead in, with a timestamp. The seven-day follow-up window depends on it.
- Re-check old consent. Legacy consents count only if obtained through verifiable means and registered. If you can't prove it, don't rely on it.
- Honour opt-outs fast. Once someone opts out, TRAI's 2025 amendment bars you from seeking consent again for 90 days.
- Run DND and calling-hour checks. Our platform applies DND checks, calling-hour windows and India-only number filters before dialling; whichever system you use, make sure these run automatically.
Common mistakes with the new TRAI rules
The most common mistake is assuming the rules target only robocall spammers. Others: treating an AI agent as "not automated" because it sounds human, calling month-old enquiries as if they were fresh, relying on undocumented consent, and leaving declaration to someone else without checking it was done.
- "Our agent sounds human, so it isn't a robocall." The A2P definition turns on how the call is started, not how natural the voice sounds.
- Treating every lead as permanently warm. The inquiry follow-up window is seven days.
- Bulk-dialling old lists from new SIMs. This is the pattern AI detection is built to catch, and penalties reach every telecom resource you hold.
- Assuming the termination charge won't touch you. How it is passed on isn't set out yet. Ask your provider how it will show up in pricing; our explainer on AI calling pricing models covers the other cost factors.
For the wider legal picture, including DND, the DPDP Act and sector rules for lending and insurance, read our guide to whether AI calling is legal in India. For practical steps on numbers and pacing, see how to avoid the spam flag on business calls, and if you lend or collect, our page on AI calling for NBFCs and lenders.
This article is general information, not legal advice; check the gazetted text and take advice for your situation. TRAI rules for AI calls reward businesses that call people who asked to hear from them, promptly and transparently. If you want to see how a compliant outbound AI calling set-up works, book a 20-minute demo and we'll call your phone with an agent set up for your business.
Frequently asked questions
Do TRAI rules apply to AI voice agent calls?
TRAI's 2026 amendment defines A2P calls as voice calls started by an application, software system or automated platform without direct human dialling, including artificial-voice calls. An AI agent that dials leads automatically fits that description. The amendment, as reported, doesn't name AI voice agents specifically, so confirm your set-up with your telecom provider.
What is an A2P call under TRAI's rules?
An application-to-person (A2P) call is a voice call initiated by an application, software system or automated platform without direct human dialling. TRAI's definition includes autodialled calls, robocalls and pre-recorded or artificial-voice calls. Senders must declare A2P use and the numbers they will use to their telecom provider in advance.
What are 140 and 1600 series numbers?
The 140 series is for regulated promotional calls. The 1600 series (1600xx and 1601xx) is for service and transactional calls, allotted by DoT to regulated entities such as banks, insurers, mutual funds and government bodies. Under TRAI's 2026 amendment, call-management apps may not blanket-block or spam-tag these series.
When do TRAI's 2026 anti-spam rules take effect?
TRAI announced the Third Amendment on 18 September 2026. Legal commentary reports a phased start counted from publication in the Gazette: general provisions after 30 days, A2P declarations after 60 days, and the appeal mechanism and multi-operator AI matching after 90 days. Check the gazetted text for exact dates.
What happens if a business makes undeclared automated calls?
Under the 2026 amendment, A2P calls made without the required prior declaration are treated as unsolicited commercial communication (spam). That exposes the sender's numbers to complaints, AI-based flagging and graded action, from KYC re-verification and a 15-day bar on outgoing services up to one-year disconnection and blacklisting for repeat violations.
See it on your own leads
In a 20-minute demo we set up an agent for your business and call your phone while you watch.