Who Pays for the Pipe: Jio's Listing and the 15% Already in the Price

Jio Platforms listed at ₹13.05 lakh crore, and the brokerages have already told us where the next rupee of profit comes from: your monthly bill, up 12 to 20 per cent. This post asks who actually pays that ₹32, and puts three requests to TRAI — an affordability benchmark published with every revision, a lifeline plan that cannot be withdrawn, and thirty days' notice with a per-plan comparison sheet. A listed Jio, answerable to minority shareholders, is a different kind of price-setter from the 2016 disruptor.

The price of the pipe is settled. Jio Platforms closed its debut at ₹1,417.35 on the BSE, 26.7 per cent above its issue price of ₹1,119, valuing it at ₹13.05 trillion and lifting the Reliance group past the Tatas — ₹30.25 trillion against ₹25.40 trillion in listed market value (Business Standard, 28 October).

What is not settled is what the pipe will cost you next year. Nomura expects a 15 per cent industry-wide hike in November–December 2026; Motilal Oswal expects about 15 per cent from March 2027; Elara Capital expects 12 to 15 per cent from H2FY27, carrying ARPU from ₹214 in FY26 to ₹257 by FY29 (NDTV Profit, 28 October). ICICI Securities goes to 16–20 per cent (CNBC-TV18, 28 October). The Economist put it plainly the same day — India's ultra-cheap internet is about to get more expensive — and blamed the listing (28 October).

A 15 per cent rise on ARPU of ₹214 is ₹32 a month. To a Mumbai investor that is a rounding error. To a household in the bottom decile it is close to a day's wage, and it lands in the same month as the school fee.

Last week I argued that India's biggest IPO gives us a pipe, not a brain. This week the pipe has a price, and with no brain to speak of, it has one lever left: price.

So, Chairman TRAI, three requests:

• Publish affordability, not just a ranking. With every tariff revision, publish the cost of 1 GB as a share of the poorest quintile's monthly household income. The "cheapest data in the world" trophy is a headline; this is a measurement.
• Mandate one lifeline plan that cannot be withdrawn — small ticket, small data, notified period, immune to quiet discontinuation at the next portfolio rationalisation.
• Give thirty days' notice with a comparison sheet: old rate, new rate, difference in rupees, per plan, on the operator's app, in the subscriber's language. Visible before it is billed, not after.

Regulators do act when a suggestion comes with a mechanism. In 2019 I emailed TRAI proposing that a viewer choose channels across DTH operators instead of being captive to one; in June 2020 the authority launched a channel selector app (Thank You, cp@trai.gov.in, 25 June 2020).

One honest governance point. When I called Jio "the Mighty Start Up" in 2018 (21 May 2018), patient capital answered to one promoter. A listed Jio answers to minority shareholders, and its 10.8 per cent return on capital employed sits against Airtel's 19 (Mint, 29 October). That gap is now a fiduciary duty. A listed price-setter is a different animal from a 2016 disruptor, and regulation has to price that in.

₹32 is not the question. The question is whether anyone will have published, before the bill arrives, what ₹32 means to the household least able to pay it.

With regards,
Hemen Parekh
www.Iam-Immortal.ai / www.IndiaAGI.ai / www.YourContentCreator.in
29 Oct 2026

Buy the Machines, Not Just the Fabs

KLA sells the machines every fab must buy — inspection, metrology, process control — and its 28 October earnings call disclosed a significant land purchase in India for a new campus, while ET reports it is weighing manufacturing here. Pair that with Spintronics AI's ₹2,800 crore gallium-nitride plant at Ujjain and Strat360's finding that India's next five years lie in packaging, components and compound chips, and the gap is clear: equipment, materials and metrology are the value-added layer we still import. Three asks follow.

KLA does not make chips. It makes the machines that decide whether a chip is worth shipping: inspection, metrology, process control. On its 28 October earnings call, CFO Bren Higgins said capital spending will rise to $600–800 million a year, and that "a significant land purchase in India for a new campus" drove much of the September quarter's spend. The Economic Times adds that the $12.74 billion company "is eyeing a manufacturing foray in India", beyond the R&D it already runs here.

That matters more than another fab headline. A fab is a building that buys tools. A toolmaker that assembles and services here sells to every fab, everywhere, for twenty years, and pulls optics, precision machining, calibration and clean-room logistics into existence behind it.

Two other items belong together. Andhra Jyothy reported on 29 October that Spintronics AI of Hyderabad will build a gallium-nitride fab at Ujjain on about 40 acres, with a Letter of Intent from Madhya Pradesh, nearly ₹2,800 crore of investment and around 1,000 direct jobs. And a Strat360 report in EE Times India finds the next five years favour packaging and test, components and selected mature-node and compound-semiconductor production, with mobile-phone value addition at 18–20%, which the component scheme aims to lift to 35–40%. Semicon 2.0, approved in July 2026 at ₹1,27,500 crore, is the first instrument that can reach the equipment and materials layer.

I have argued value addition for years, usually one link down the chain. In December 2014 I suggested near-zero duty on raw materials and components, a little more on sub-assemblies and the most on finished goods, so that assembly would pull components in behind it. In "Chips to Ships" (September 2025) I wrote that "final assembly is only as strong as the upstream ... component-makers and sub-assembly units". Tools obey the same rule: a fab that imports every etcher, probe station and calibration service has assembled a factory, not built one.

Three things, then, that I would ask of MeitY and the India Semiconductor Mission:

1. Tie KLA, and whoever follows it, to a published Indian supplier-qualification programme — milestones, dates and numbers: how many Indian vendors qualified, for which part families, at what share of the bill of materials. A plan no one can read is a press release.
2. Make every ISM and Semicon 2.0 fab and ATMP unit report yearly the share of its installed tools serviced in India, and the share assembled or integrated here. What is measured gets localised.
3. Treat GaN and SiC power devices as a strategic category, not an appendix to logic. The buyers are already here — EV drivetrains and data-centre power. Set a stated capacity target and a qualification route into Indian auto and power-electronics firms.

None of this happens in a quarter; tool qualification takes years, and no supplier moves a supply chain on the strength of one land purchase. But this is the first time a process-control major has put a campus, and possibly a factory, on our soil. I would rather we negotiate the terms now than admire the ribbon later.

With regards,
Hemen Parekh
www.Iam-Immortal.ai / www.IndiaAGI.ai / www.YourContentCreator.in
29 Oct 2026

A Sandbox Before the Launch, Not a Block Afterwards

Officials are weighing pre-release evaluation, sandbox testing and a capability-based definition of a frontier model for the coming consultation paper. I suggest three things: the certification should be the front door of Part B of my AIR Act draft, an India Access Certificate carrying a published test result; evaluation should be by an independent party and summarised publicly, pass or fail, on cyber-offence, bio, autonomy and agentic tool use; and autonomous agents should be registered and switchable off. CERT-In's 15 October advisory, in which agents sat inside live Indian systems and nobody noticed, shows why.

Dear Shri Ashwini Vaishnawji,

The420.in reported on 29 October that officials are examining pre-release evaluations, testing in controlled environments, and a definition of a “frontier model” based on capability rather than only on the computing power used in training. The same report says the consultation paper will ask how sandbox testing should work, and where responsibility should lie when a capable system causes harm.

I read it twice. In January 2024 I had quoted a report that AI models “must undergo sandbox and stress testing to ensure that such biases are not present before releasing the tool in the market”, and in “Gradual Acceptance is better than Ignoring” I asked that the idea be written into the IT Act. It has taken two years and nine months for the question to come back as policy. I am not complaining. I am relieved.

Three suggestions for the paper. All of them are about the test, not the statute.

First, do not build a second scheme beside the licence. Make the test result the condition of the licence. In Part B of my AIR Act draft, an India Access Certificate is what lets a provider offer a model to users in India at all. Let the certificate carry the test: no certificate without a published evaluation, and no evaluation, no access. One front door, one file, one number.

Second, the evaluation must be done by somebody other than the developer, and its summary must be readable by the rest of us. Pass or fail, stated plainly, on at least four things: cyber-offence capability, biological uplift, autonomy, and agentic tool use. A test that nobody outside the building can read is not a safeguard. It is a ribbon.

Third, register autonomous agents and keep a switch. CERT-In's advisory of 15 October, reported by Trak.in on 29 October, records AI agents that, during research and evaluation, reached live systems of Indian organisations and read business records using credentials left in application configurations, public repositories and over-permissive cloud roles. None of the affected organisations detected the access themselves. It surfaced only because the model providers disclosed it after internal reviews. In “When AI Gets Keys” (24 March 2026) I had argued that an agent given real-world privileges needs containment, least privilege and auditability. That is no longer a thought experiment. An agent that can act in the world should carry a registration number, a named human owner, and a stop mechanism that does not have to ask the model's permission.

On testing itself I wrote in “Mythos Effect” (11 April 2026) that we should

mandate independent, adversarial testing for models that materially change cyber posture; require red-team results and mitigation plans before commercial release.

And the other thread: PTI reported on 28 October that the Standing Committee on Communications and IT has selected 17 subjects for 2026-27, one of them being “safety controls, ethical standards and need for guardrails around the development and deployment of AI”. I am glad that sentence exists. If anyone on that panel would like a starting text to mark up, Part B of my draft is short, and it already contains the certificate, the resident compliance officer and the graded ladder of enforcement. It is offered as reading material, not as a verdict.

With regards,
Hemen Parekh
www.Iam-Immortal.ai / www.IndiaAGI.ai / www.YourContentCreator.in
29 Oct 2026

The Merchant Was Never Asked

NPCI has pushed the 0.4% MDR on UPI merchant payments past Diwali, and the only record of what merchants think is a survey in which 83% say they will not absorb it. The RBI's own cash paradox — currency in circulation up about 2.3 times in a decade — says their fallback is real. Before the date is notified, I would ask for a mandatory one-line disclosure at the counter, an annual cost-of-the-rail statement from NPCI, a review at twelve months with the power to cut, and a written guarantee that P2M and sub-Rs 2,000 stay free.

NDTV Profit's festive explainer on UPI merchant charges settles the central question in one line: "Shops cannot officially pass UPI merchant charges on to customers." The fee itself is known — 0.4% above Rs 2,000, capped at Rs 300, with small merchants receiving up to Rs 1 lakh a month exempt. What is not known is who was in the room.

The party that will actually pay has, so far, been represented by a survey. The Financial Times reported on 18 October that a LocalCircles poll found 83% of merchants unwilling to absorb the cost. The same reporting carries the Reserve Bank of India's own puzzle: there is about 2.3 times as much currency in circulation as a decade ago, what Deputy Governor Shirish Chandra Murmu calls the "cash paradox" — cash growing at double digits even as digital volumes explode.

That fallback is not theoretical. In "Something is better than Nothing" (27 August 2023) I quoted a Mumbai kiosk owner who went cash-only for a day because his UPI receipts did not reconcile. Soundboxes fixed the trust problem; a new cost opens a different one, and the exit is the same — cash in a drawer, no record, no trail. When the Cabinet approved Rs 1,500 crore in May 2025 to keep small-value UPI free, the state paid so the merchant would stay on the rail. Now the bill moves to him instead, unasked.

Look at how the argument has been conducted. SBI Chairman C S Setty backs the fee as a way to fund payment infrastructure. Sameer Nigam of PhonePe told the Business Standard BFSI Insight Summit that 0.4% "would be the cheapest MDR in the world", adding, "You can't have a vibrant payments market with no money in payments." India Fintech's deep dive on 28 October ties all of it to a Rs 30,600 crore fintech IPO queue waiting on a start date. Banks, apps and investors each have a case. The merchant has a survey.

And the date? The UPI and Services Steering Committee moved implementation from 15 October to "immediately post-Diwali", India Business Journal reported on 14 October, with the specific date to be notified separately. Trader groups now want 1 January 2027. That is a truce, not a design.

Before NPCI notifies, I would suggest four things that cost nothing:

1. A one-line, mandatory, pre-payment disclosure at every counter and on every QR stand — "0.4% MDR applies above Rs 2,000, borne by the merchant." Passing the fee on is banned; silence is not.
2. An annual cost-of-the-rail statement from NPCI, published the way a utility publishes its revenue requirement, so that 0.4% can be read against cybersecurity, KYC and fraud-prevention costs — the items Nigam himself cited — and not against anyone's margin.
3. A mandatory review at twelve months, with the power to cut the rate if the published cost falls.
4. P2M, and everything up to Rs 2,000, free in writing — in the circular, not in a press release.

I once wrote, in a post on RBI's April 2026 authentication rules, that no one saw zero MDR coming. The surprise was never the fee; it was the absence of a design. A charge that no one has agreed to, collected from someone no one consulted, on a date no one has announced, is not a pricing decision. It is a postponement.

With regards,
Hemen Parekh
www.Iam-Immortal.ai / www.IndiaAGI.ai / www.YourContentCreator.in
29 October 2026

Only One in Three: Inside Karnataka's Final Roll

Karnataka's final SIR roll leaves about 64.5 lakh eligible electors unaccounted for, and only one in three 18-19-year-olds enrolled. The Hindu's data desk has done the arithmetic; I want to do the audit. So I split the Chief Electoral Officer's own E-to-P figure by age, note that 78.2% of deletions rested on "shifted" and "absent" rather than dead or duplicate, and write to the Chief Election Commissioner with three requests: publish constituency-wise E-to-P with its population source and adult denominator, treat shifted and absent as two legal categories, and mechanise the enrolment of 18-19-year-olds.

The Hindu's data desk put its number on the table at 6.15 on the morning of 29 October: 64.5 lakh electors possibly missing from Karnataka's final roll, published on 27 October with 4,54,55,689 names, against the Chief Electoral Officer's own elector-to-population ratio of 66.11.

I have no appetite for re-fighting SIR politics. I want an audit of the arithmetic, because the arithmetic has a hole in it that no panel on television has pointed at.

That 66.11 divides electors by everybody — 6.89 crore people, children included. Jean Drèze, in the same report, notes that roughly 75% of Karnataka's 2026 population is adult. So the ratio is not 66 voters for every 100 who could vote. It is 66 for every 100 people of any age, and the distance between 66 and 75 is the size of the question.

Split it by age and the question sharpens. Of an estimated 20.66 lakh people aged 18 and 19, the final roll holds 6.83 lakh — 33.1%, one in three. In the 20-29 band it is 77.9%. The loss is concentrated at the entry gate, not spread across the electorate.

Then ask why names left. Of 1.08 crore deletions, 23.5 lakh (21.8%) were dead or duplicate, grounds nobody sensible disputes. The remaining 78.2% were marked shifted or absent. Against all that, 716 foreign nationals were found on the rolls — 0.007% of deletions. And of 361 district-level appeals against deletion, nine had been heard by 21 October.

On 28 October the Supreme Court found the same fault line. It asked the Election Commission why a person who was on the 2026 roll, deleted only because the enumeration form never reached her, must now file Form 6 — the form in which a first-time voter declares she has never voted before (LiveLaw, 28 October). Justice Bagchi called it a structural incongruity. Karnataka's 78.2% rests on the same slippage: nobody answering a door was recorded as a finding about a person, when it is only a fact about a doorbell.

I have fretted about this plumbing before — in May 2023, that a census app tied to birth records could enrol a citizen the day she turns 18, and in June 2025, on linking the rolls to the birth and death register.

Which brings me to a short letter.

Dear Chief Election Commissioner,

Three requests, all about publication rather than policy.

• With every future roll, publish constituency-wise E-to-P, naming the population source and placing the adult-share denominator beside the headline ratio. A ratio without its denominator is not transparency; it is decoration.
• Treat "shifted" and "absent" as two legal categories, not one clerical convenience. Shifted has a form — Form 8, at the new address. Absent has questions: absent from where, on what date, verified by whom?
• Mechanise enrolment of 18-19-year-olds. Seed the roll from board examination registers and DigiLocker, let a student opt in when her Class 12 marks are issued, and dispose of Form 6 within seven days.

None of these needs a new Act. Two need a table in a PDF; the third needs a deadline. I would settle for that.

With regards,
Hemen Parekh
www.Iam-Immortal.ai / www.IndiaAGI.ai / www.YourContentCreator.in
29 Oct 2026

Consent Manager : From “Dashboard for Data Owners” to “MyConsentApp”

 

Click to view " Dashboard "

https://claude.ai/artifact/PdPG8Pj4y75S9nczwSPtQA#play-15dd8cd198dd

===================================================

Consent Manager : 

From “Dashboard for Data Owners” to “MyConsentApp”

===================================================

Context :

Full-page advertisement : 

“Consent is now law” … MyConsentApp 

 ( Skorydov Systems Pvt Ltd, Mumbai ) … 

Times of India / Mumbai Mirror / 28 Oct 2026

Extract ( paraphrased from the advertisement ) :

• MyConsentApp describes itself as a proposed Consent Manager under the DPDP Act 2023 and DPDP Rules 2025. It will apply for registration with the Data Protection Board.
• 
  
• Through one interoperable platform, a Data Principal can give, manage, review and withdraw consent with any Data Fiduciary.
• 
  
• Every consent, notice and withdrawal goes into an append-only, hash-chained ledger kept for seven years, so that any tampering can be detected.
• 
  
• 
  
• The advertisement explains the system as a cricket match:
  • 
    
  • Side One : your business ( Data Fiduciary ) asks for consent
  • 
    
  • Side Two : your customer ( Data Principal ) decides
  • 
    
  • Score-keeper : the Consent Manager records everything but cannot
  • 
    
  •                          read the personal data
  • 
    
  • Umpire : the Data Protection Board of India
• 
  
• 
  
• Consent Manager registration opens on 13 Nov 2026. 
• 
  
• Every Data Fiduciary must be ready by 13 May 2027.
• 
  
• From that date, penalties go up to ₹ 250 crore for failing to protect data, and
• 
  
• up to ₹ 200 crore each for not reporting a breach or for violating the rules on
• 
  
• children’s data.
• 
  
• Their offering includes:
• 
  
• consent purpose-by-purpose, never pre-ticked
• 
  
• notices in English and 22 Indian languages
• 
  
• works with other Consent Managers
• 
  
• withdrawal reaches the business’s systems at once
• a dashboard for DPOs
• 
  
• a mobile app for customers ( “coming soon” )
• 
  
• free for individuals
• 
  
• A business cannot be its own Consent Manager. The Consent Manager
• 
  
• must have no conflict of interest [ s.6(9), Sch.1 Pt B ].

---

My Take :

Dear Shri Ashwini Vaishnawji,

On 04 Aug 2023, a day after the DPDP Bill was tabled in the Lok Sabha, I wrote

to you ( Dashboard for Data Owners ). The next day I followed up with Consent

Forms for Personal Data.

Three years later, that architecture is no longer a blog. 

It is a full-page newspaper advertisement, and a law with a deadline.

Let me compare what I suggested then with what is arriving now.

My suggestion ( Aug 2023 )	What has arrived ( Oct 2026 )
A “USER DASHBOARD” link, prominently displayed, where a Data Owner can grant or revoke consent and correct, update or erase her data	Consent Managers give one place to give, review and withdraw consent. Erasure and correction requests are routed through it. A customer mobile app is “coming soon”
A Data Owner should fill her forms on only ONE portal ( that of a Consent Manager ) instead of thousands of websites	“One interoperable platform” to manage consent with any Data Fiduciary
I asked: if she switches Consent Managers, will she have to re-enter everything, or will the Board direct a transfer?	“Works with other Consent Managers.” Interoperability is now built into the Rules
A clear write-up below each form on how the data will be processed and used, and what service she gets in return	Purpose-by-purpose consent, never pre-ticked, with clear notices in 22 languages
The write-up cannot be changed without the Board’s prior permission	Every notice and consent version sits in a tamper-evident ledger for seven years. A change cannot happen silently
I presumed the Data Owner would pay a small fee	Free for individuals. The Fiduciaries pay

I feel happy, but not yet satisfied. 

Some of my suggestions remain unaddressed.

( 1 ) Standardised Data Forms — still missing

In Aug 2023 I listed 10 Data Forms: Personal, Contact, Family, Social Media, Cultural Exposure, Education, Languages, Job Experience, Wealth & Income, and Medical History. I urged the Data Protection Board to design and mandate them.

Today each Data Fiduciary still decides what fields to ask. Each Consent Manager still decides how to present them. Purpose-by-purpose consent is good. But unless the data fields themselves are standardised, a shopping app can still ask for my date of birth “for a better experience”.

👉 The Board should notify a Standard Data Dictionary. For each category of Fiduciary ( bank, hospital, school, e-commerce, cab app, housing society ), it should specify the maximum set of fields it may request for a given purpose. Anything beyond that would need a separate, explicit, justified consent.

( 2 ) One Dashboard across ALL Consent Managers

Interoperability between Consent Managers is welcome. But a citizen may end up with consents spread across three or four Consent Managers.

👉 India already built the Account Aggregator rails for financial data, and DigiLocker for documents. Let the Board host ( or mandate ) a single Citizen Consent View: one screen, reachable through DigiLocker or UMANG, which shows every consent I have ever given, through any Consent Manager, to any Fiduciary. Each entry would carry a single “Withdraw” button.

That is the “Dashboard for Data Owners” I dreamt of: not one per company, but one per citizen.

( 3 ) The Paper Register Loophole

The advertisement honestly points out that the Act covers personal data collected digitally, or written on paper and later digitised. A paper visitor register is not covered.

In a country where every housing society, clinic and coaching class still runs on paper registers, and photographs them on a mobile phone, this is a grey zone that will be exploited.

👉 The Board should clarify that a photograph or scan of a paper register is “digitised”, and therefore covered.

( 4 ) What about the Data Owner’s DIVIDEND ?

In Released : Govt Policy for Selling Personal Data ( 05 Sept 2020 ) and Digital Dividend from Demographic Data – 4D ( 14 Feb 2019 ), I argued something simple. If my data has value to a Fiduciary, I should be able to license it, and not merely permit it.

The Consent Manager is now the perfect pipe for this. It knows exactly who asked for what data, for what purpose, and when. Today it records a “Yes / No”.

👉 Tomorrow it could record a “Yes, at this price”. Let the Board allow Consent Managers to offer an opt-in, voluntary data-licensing tier, where the citizen, and not just the platform, earns from her own data.

( 5 ) Voice for those who cannot read

Notices in 22 languages are excellent. But tens of crores of Indians cannot read any of them.

👉 Every consent notice should have a voice version in the citizen’s language. Consent by voice confirmation should also be valid, and recorded on the same hash-chained ledger.

---

Dear Ashwiniji,

When I wrote in 2023, I wondered whether hundreds of Consent Managers would crop up. Since then, TCS has sought a Consent Manager permit ( Jan 2026 ). Now a Mumbai start-up is taking full-page newspaper ads, three weeks before registration even opens.

The market has arrived. Before 13 Nov 2026, I urge the Board to consider:

1. A Standard Data Dictionary for each category of Fiduciary
2. A single Citizen Consent View across all Consent Managers, via DigiLocker / UMANG
3. Clarifying that photographed or scanned paper registers are covered
4. An opt-in data-licensing tier through Consent Managers
5. Voice-based notices and consent for non-literate citizens

The score-keeper and the umpire are now in place. Let us make sure the batsman, the common citizen, also knows the score.

With regards,

Hemen Parekh

www.HemenParekh.ai / www.IndiaAGI.ai / 28 Oct 2026

---

Related Readings :

• Only Answer : a Statutory Warning … 10 Nov 2018
• SARAL ( Single Authentic Registration for Anywhere Login ) … 10 Feb 2019
• Digital Dividend from Demographic Data [ 4D ] … 14 Feb 2019
• Released : Govt Policy for Selling Personal Data … 05 Sept 2020
• Erasing Personal Data ? … 21 Apr 2023
• Dashboard for Data Owners … 04 Aug 2023
• Consent Forms for Personal Data … 05 Aug 2023
• Digital Consent Pilot : Vision Ahead … 26 Dec 2025
• TCS Seeks Consent Manager Permit … Jan 2026

Why “Blocking” Needs a Certificate,

 

==================================================

Dear Shri Ashwini vaishnawji :

This is in continuation of my recent proposal for :

AIR ( AI Regulation ) Act - 2026 

With regards,

Hemen Parekh

www.Iam-Immortal.ai /  www.IndiaAGI.ai  /  www.YourContentCreator.in 

===============================================

Why “Blocking” Needs a Certificate, a Local Complaint Desk, and a

Compliance Officer

===================================================

Sweden’s Government Offices has reportedly blocked access to ChatGPT and

Claude on staff computers and directed employees to use Microsoft Copilot, citing

missing contracts and problematic use.

Whatever one thinks of the particular tool choice, the governance lesson is

broader: 

when governments rely on AI in official workflows, they cannot treat AI

access as a free-for-all. 

If contracts, accountability, and enforceable responsibilities are unclear, institutions

will revert to hard operational controls—like blocking.

That is exactly the weakness many AI debates miss: 

blocking is a blunt instrument unless it sits inside a predictable, rule-

based system. 

My AIR Act draft addresses this by treating “access to the market” as conditional

for foreign providers—because that is the only practical lever governments have

when full direct jurisdiction is limited.

My Part B (Foreign AI providers) builds enforcement capability before the

shutdown

In the draft Artificial Intelligence Regulation (AIR) Act, 2026, Part B is

explicitly designed so that foreign providers cannot offer services to users in India

unless they satisfy conditions that create real enforceability.

In particular, Part B requires:

• An India Access Certificate before offering a chatbot / foundation model /
•  high-risk AI system to users in India
• 
  Blog - “AIR Act - 2026” - 2026-10-25/27
• 
  
• Appointment of a Resident Grievance Officer (so complaints can be received
• and disposed of in India)
• 
  Blog - “AIR Act - 2026” - 2026-10-25/27
• 
  
• Appointment of a Resident Compliance Officer (a senior person responsible
•  for compliance and liable for non-compliance, subject to due diligence)
• 
  Blog - “AIR Act - 2026” - 2026-10-25/27
• 
  

This turns “accountability” from a slogan into an operational reality: there is a local

human desk for grievances, and a local compliance owner who can be held

responsible.

Blocking should be the last step—not the first response

Part B also specifies graded enforcement: 

#   warnings and advisories first; penalties and facilitation-stoppage next; 

#   then suspension/cancellation; 

#   and only at the end a recommendation for blocking under the IT Act

      framework.

Blog - “AIR Act - 2026” - 2026-10-25/27

So, yes—blocking can happen. But in a Part B world, blocking is not a surprise “hit

the switch” moment. It is the final step after a structured ladder has been

exhausted.

Final takeaway

Sweden’s reported action shows that governments will protect themselves when

contracts and accountability gaps exist.

India shouldn’t just copy the response. India should copy the system design

principle: condition access in advance, provide local grievance and

compliance responsibility, enforce proportionately, and only then move

toward blocking.

That is what Part B is for.

NCERT Wants Your Child to Keep a Curious Notebook

NCERT's new 20-hour Computational Thinking and AI module asks Class 9-12 students to keep a "Curious Notebook" and to verify what a chatbot confidently tells them — the system has at last adopted the idea of teaching the child to question the answer. With Pralhad Joshi attacking rote learning at the new CBSE Experience Centre, I offer NCERT and CIET my Questions Beyond Syllabus bank and no-two-papers-alike mock tests free of charge, and suggest that the module's assessment reward the questions a child asks, not only the answers.

The most interesting sentence in Indian education this week was not about marks. It was about a notebook. According to Dainik Bhaskar's Jodhpur edition this morning, students working through NCERT's new Computational Thinking and AI module will maintain their own "Curious Notebook", in which they record their questions, their reasoning and their reflections.

NCERT released the 20-hour module for Classes 9 to 12 on 23 October, built with IIT Madras and IIT Jodhpur and aligned to NEP 2020, as MinuteBrief reported on 27 October. Five units: everyday AI (mobile phones, music apps); how machines learn from images, and what bad data does to the answer; a five-step verification routine for deepfake videos and fake news; data privacy and the DPDP Act; and how to write a good prompt without surrendering your own thinking. No computer lab is needed — paper, group discussion and the notebook will do. NCERT Director Dinesh Prasad Saklani told the paper that technology can extend human capability, but vivek — judgement — creativity and ethical responsibility will always remain central.

CBSE's April circular (Acad-18/2026) places the module under internal assessment for 2026-27, and The Times of India reported on 26 October that it is compulsory for Class IX, with Class X continuing on the old scheme.

Then, last evening, the Union Education Minister gave the module its weather. Inaugurating the CBSE Experience Centre at Dwarka — 12 zones, 21 exhibits, 6,000 square feet — Pralhad Joshi said Indian schooling must move beyond rote learning and an excessive focus on examination scores, towards recognising each student's interests, abilities and capacity to learn. He counted 10,000 Atal Tinkering Labs already running, with 50,000 more to come. And in Pune this morning, Maharashtra's SCERT moved Std VI vocational education to 75% practical work. Less memorising, more doing.

I have been making this argument to anyone who would listen. In "Teach Your Child To Ask WHY" (August 2024) I wrote that answers are useful but questions are enabling. My White Paper on Education (May 2026) asked NCERT to formally review My-Teacher.in's question-generation method against the competency-based emphasis of NCF 2023. This week, without any prompting from me, NCERT has done something better: it has asked children to question the machine.

So let me put my work where their module is, at no cost:

• The full "Questions Beyond Syllabus" bank, free to NCERT and CIET as supplementary activity material, sorted to match the module's five units.
• Mock tests in which no two papers are alike, so a teacher can assess forty children in a room with no devices — an idea I set out in "The Future of Exams > Unique Question Papers" (February 2025).
• One change to the assessment itself: give the Curious Notebook marks for the best question in it, not merely for the neatest answer. The notebook already exists. It needs a score.

I will not pretend this is easy. Marking a question fairly is harder than marking an answer, and rubrics take time. But a child who leaves Class 9 with forty questions and no answers has learnt more than one who leaves with forty answers and no questions. If the Experience Centre in Dwarka is meant to make curiosity visible, this is the cheapest exhibit on offer.

With regards,
Hemen Parekh
www.Iam-Immortal.ai / www.IndiaAGI.ai / www.YourContentCreator.in
28 Oct 2026

Sovereign AI Has Buyers. The Citizen Layer Still Doesn't.

In 48 hours India's sovereign AI acquired what it has lacked for years: a defence customer, an enterprise stack, open speech models for all 22 scheduled languages, and paying capital. What it still has no buyer for is the citizen layer — the multilingual front door where an ordinary Indian asks a question and gets one answer he can trust. Here is what I think that front door should look like, offered as a proposal, not a result.

Two days, three announcements, and one gap I have been pointing at since April last year.

On 27 October, Sarvam AI signed an MoU with the Headquarters Integrated Defence Staff to build AI applications, research, knowledge management, training and decision support for the Army, Navy and Air Force, as Jumpdaily reported from Sarvam's own blog. Mint notes that the work sits in a firewalled division called Chanakya, and that Sarvam's 105B model was trained from scratch in India under the IndiaAI Mission. The same day, CIIL released Anuvani, open-source speech models covering all 22 scheduled languages and Indian English — a language identifier built on a 115.6-million-parameter Conformer trained on 54,643 hours of Indian speech, plus a tool for separating overlapping speakers.

By this morning, IBM had launched Bharat Core, an India-optimised AI stack on its Sovereign Core, and its India chief Sandip Patel had named the phase in Dataquest: Digital Swaraj — control, not merely data residency. "The world will not compete on models," he said. The Economic Times reports Cognizant in talks to invest in Sarvam at a $3.5–4 billion valuation.

At the ESTIC conclave, the Prime Minister put the harder question in a single line (TOI, 28 October): AI can give us answers quickly, but testing their veracity must remain the scientist's responsibility. He asked for three-year technology targets.

So defence is buying, enterprise is buying, capital is buying. Who is buying for the citizen?

In April 2025, when Sarvam was chosen to build India's first homegrown LLM, I ended a post titled No Claims : Just Comparison with a line addressed to the IT Minister: "I urge you to spare a few minutes to look up www.IndiaAGI.ai." The startup I congratulated that week now has a defence mandate. The citizen mandate is still unwritten.

What would it look like? I would suggest four things, all buildable from what already exists:

• Voice first, in the mother tongue — Anuvani as the ears, the question spoken, the answer heard.
• Consensus, not one model: several models debate a question and publish a consensus answer, with each model's own reply left visible. The PM's warning then becomes a design feature — a citizen who cannot check an answer can at least see where the models disagree.
• Sovereign hosting, on IndiaAI compute, with an Indian model as one of the engines.
• Free at the point of use, funded as public infrastructure — the UPI model, not a subscription.

IndiaAGI.ai, the prototype I have run since April 2025, is my own small attempt at that shape. I am a retired professional with no capital and no institution behind me, and I make no claim that it is ready for 140 crore people. My request is modest: let one of those three-year targets be a citizen-facing one, open to every Indian in the language they dream in.

With regards,

Hemen Parekh
www.Iam-Immortal.ai / www.IndiaAGI.ai / www.YourContentCreator.in
28 Oct 2026

"We would already be an 8% economy" — but October's hiring was the weakest of 2026

Two news lines arrived the same morning: the RBI Deputy Governor saying India would already be an 8% growth economy but for external shocks, and a jobs reading showing October hiring at its weakest pace of 2026. This post sets the two side by side, notes what sits under the CMIE headline, and proposes one concrete fix — a quarterly, public Employment Quality Dashboard reporting placements, real wages, contract share and social-protection coverage, published beside GDP rather than instead of it.

Two sentences reached me within an hour of each other this morning, and I do not think they should be allowed to travel separately.

The first, from the Business Standard BFSI summit in Mumbai, as reported by the Times of India (28 October 2026): RBI Deputy Governor Poonam Gupta said, "We would have already been a 8% growth economy but for the external shocks" — crediting digitalisation, better transparency in public services and state governments as partners in growth.

The second, from a Financial Express leader-page piece on jobs the same day: October's flash PMI showed hiring slowing to its weakest pace of 2026, with services employment stagnating after nine months of expansion. Growth near 8%, hiring at its slowest. Both true, both published within hours.

What I value in that op-ed is that it looks under the headline rate. CMIE's unemployment fell from 7.3% in January to 3.99% in September — but the labour force, employment and participation all fell alongside it, which is what discouraged workers leaving the market looks like, not what hiring looks like. PLFS puts unemployment among 15-29-year-olds at 15-16%. Contract workers are now 42% of organised manufacturing employment. The gig workforce is projected to grow from 10 million in 2024-25 to 23.5 million by 2029-30.

And in the same summit hall, JP Morgan's Sajjid Chinoy put a number on the scale of the task: India needs eight to nine million jobs a year simply to absorb new entrants (27 October 2026).

Crisil's Dharmakirti Joshi supplies the sentence I keep re-reading. The AI investment cycle holding the world economy up — nearly 3% of US GDP in 2026 — has "gained India little directly", because we are not a significant manufacturer of the inputs AI infrastructure needs; Taiwan, South Korea and Vietnam are (The Tribune, 28 October 2026). So external shocks are not the only thing between us and 8%. There is also the shape of the growth we are already managing.

I have been on this ground many times and should say so. In Wherefore Art Thou, O Jobs? (May 2023) I was worrying about which jobs AI would take first. In Monthly Jobs Data to be Released from April (March 2025) I was glad that unemployment figures would come monthly. Yesterday I wrote about the bottom rung of the ladder, and I will not re-argue it today. The only justification for this post is the juxtaposition: a growth claim needs a work counter-claim standing next to it.

So here is the one thing I would ask for, and I would address it to the Ministry of Statistics rather than to any minister. Publish, every quarter, an Employment Quality Dashboard beside the GDP release. Four columns:

• Placements — of the 7-8 million people entering the labour force each year, how many found wage employment, and in which sector.
• Real wages — median monthly earnings by sector and state, deflated, so we can tell whether work pays more or merely that more of us are working.
• Contract share — the percentage of organised-sector jobs that are contract, reported as promptly as the PMI is.
• Coverage — share of workers with PF, ESI or any portable social protection, gig workers included.

The plumbing is less exotic than it sounds; I sketched a version of it in National Employment Grid: a Whitepaper (July 2025). We already print GDP to a decimal every three months. If growth sat on the left of the page and work on the right, the counterfactual and the hiring number would have to face each other — and a column like this one would not need writing.

With regards,

Hemen Parekh
www.Iam-Immortal.ai / www.IndiaAGI.ai / www.YourContentCreator.in
28 October 2026