Implementation Annex · Annex to Chapter 25
An Internet That Works for India
A starting draft for the group of action-minded builders, from mission heads and industry veterans to entrepreneurs and the private sector, to refine. The numbers and steps show one workable way; the details are settled at execution.
The plan
- Data in India. A full copy of Indians' personal data is kept on servers in India, the model RBI applied to payments data in 2018. Data may still be processed abroad if the India copy is complete and open to Indian courts. No new Act is needed: Rule 13 of the DPDP Rules 2025 lets the government require Significant Data Fiduciaries to keep specified personal data in India, and every platform with 50 lakh or more Indian users is designated one.
- Indian-first, not foreign-out.
- Every phone sold in India shows a choice screen at setup, with at least one Indian option for browser, app store, maps, messaging and short video.
- Government buys any product certified to an open published spec at the published price (How the government buys). Where an Indian and a foreign certified product both claim the same order in the first 24-hour window, the Indian one wins the tie before any lottery, as the Public Procurement (Make in India) Order already allows; at every other moment the first-come, first-served rule applies to both alike.
- Messaging and social apps with 5 crore or more users let users export their contacts and posts, and message across apps through one open standard, within 1 year. This is the UPI model applied to social apps. The standard is IETF MLS (RFC 9420), which keeps end-to-end encryption across apps, and CERT-In audits each app before it connects.
- Foreign apps keep running if they obey Indian law. An app that a court finds in breach of Indian law gets 90 days to fix it. Removal follows only after that, and only once an Indian alternative has at least 1 crore monthly users and has passed a CERT-In security audit, so no user is stranded.
- Take down only what is illegal. Takedowns need a court order or a Section 69A order. Emergency 69A orders (national security or public order) lapse after 7 days unless one of 25 Designated Review Benches, statutory benches of retired High Court judges, confirms them. Every order is published within 48 hours, redacted only for live operations. A closed list of content already illegal by statute or court ruling (child sexual abuse material, non-consensual intimate images, impersonation, and financial-fraud or phishing links) is removed on notice within 24 hours, with a user appeal decided in 7 days. The Sahyog portal is used only for these routes.
- Endless short-video feeds banned, permanently, for everyone. From day 30, no app or website, Indian or foreign, may offer users in India an endless short-video feed. The ban covers the design, not the videos. A feed is banned if it has all three of these features:
- (a) short videos, up to 3 minutes long;
- (b) the next video starts by itself or on a swipe, with no end point;
- (c) the platform picks the videos, by any ranking, recommendation, trending or random method, rather than the user.
What stays open: posting and watching any lawful video; search; channels the user follows, shown newest first and ending with "You're all caught up"; links shared by friends; playlists and courses the user picks; and long videos. One rule binds every app and every user (one country, one law), so no age check is needed.
- The legal route. The ban, the enforcement powers and the review benches are set by a short Online Feed Safety Act, modelled on the Promotion and Regulation of Online Gaming Act 2025, which Parliament passed in two days in August 2025 to ban online money games for every user, enforced by blocking. If Parliament is not sitting, the President promulgates it as an ordinance (Article 123) and the Bill is passed in the next session. The Act names an Online Safety Authority to test apps and set fines, with appeal to TDSAT (the telecom appeals tribunal).
- Enforced at the app store and the ISP.
- Self-declaration, no licence: every video app with 10 lakh or more Indian users declares compliance on a public register within 30 days. Accredited private labs and CERT-In run a published test, under the Aadhaar playbook's open-spec model, on randomly chosen apps every month.
- Notice: an app found running a banned feed gets 7 days to switch it off.
- App stores: after 7 days, Google Play, the Apple App Store and every Indian app store remove the app's current version for Indian users within 24 hours.
- ISPs and telecom operators: at the same time, every ISP and mobile operator blocks the feed service through the national block list (domain-name filtering, the method already used for Section 69A orders). Every app serves any short-video feature from its own named domain. An app that hides it inside its main domain has the whole app blocked until it complies.
- Fine: 2% of the platform's Indian revenue for each month a banned feed stays live, set by the Online Safety Authority within 30 days. The duty is to not serve a banned feed to users in India, judged by SIM, payment and location signals, VPN or not.
- Unblock: the platform switches the feed off and asks for a re-test, which is done within 7 days. On a pass, it is unblocked within 24 hours.
- Promote learning, skills, entrepreneurship and STEM videos. Promotion only: no one is told what to watch.
- One free national learning library: DIKSHA (school) and SWAYAM (college) are merged into one open library, with the school curriculum (Chapter 15), the science curriculum (Chapter 15) and the skills track (Chapter 23) mapped to videos. Any app, Indian or foreign, can show its videos through an open API.
- Creator Learning Fund, ₹1,000 crore a year: pays verified educational creators for each learner who passes a short skill check after a video or course. It never pays per view or per minute watched, which can be faked.
- Verified badge: independent panels of teachers, scientists and founders check a creator's videos for accuracy within 30 days. Verified videos rank first in the library's search.
- Skill credits: passed checks count toward skills milestones (Chapter 23).
- Founders' channel: the fund of funds and startup programmes (Chapter 16) publish pitch, build and scale courses taught by Indian founders.
Cost and return
- Advertising effect (estimate). Digital advertising in India was ₹94,700 crore in 2025 (FICCI-EY). If short-video feeds carry about 20% of it, about ₹19,000 crore a year is exposed. If half or more moves to search, long video and followed-channel feeds, the net loss is about ₹7,600 to 9,500 crore a year, about 3% of the ₹2.78 lakh crore media and entertainment sector.
- Enforcement and learning push (estimate). The learning push is counted in what Chapter 1's new taxes raise, so it does not draw on the Chapter 23 subsidy saving.
| Item (estimate) | Cost (₹ crore) |
|---|---|
| Enforcement: test labs, the register and the Authority | about 50 a year |
| Enforcement: ISP blocking, using existing Section 69A block-list systems | about 20 a year extra across operators |
| Learning push: the Creator Learning Fund | 1,000 a year |
| Learning push: merging DIKSHA and SWAYAM | about 200 once |
- Data in India. Large platforms already run Indian data centres; the extra storage cost is under 1% of their Indian revenue (estimate).
- Upside for Indian apps. Choice screens and open messaging lower the cost of winning users, learning apps gain the attention freed up, and the Indian-first internet sleeve of the fund of funds (Chapter 16) backs new Indian apps.
Timeline
First 30 days:
- The Online Feed Safety Act is passed, or promulgated as an ordinance, with the three-part test, the register, the 7-day notice, the Online Safety Authority and the 25 Designated Review Benches.
- The open test spec is published and the first labs are accredited.
- The ban takes effect on day 30.
- Every blocking order since January 2026 is published in a searchable register; the 7-day confirmation rule for emergency 69A orders and the closed list of illegal content are notified under the Blocking Rules.
- Platforms with 50 lakh or more Indian users are designated Significant Data Fiduciaries, and the data-in-India rule is issued for 21 days of comment.
- The Creator Learning Fund and the badge panels open.
Within 1 year:
- Audited watch time on endless short-video feeds in India down at least 90% from the day-0 baseline.
- National learning library live with the school, science and skills curricula mapped to videos.
- Data copies in India complete; choice screens on all new phones; open messaging standard live for apps with 5 crore or more users.
Guardrails
- Exact test. A feed is banned only if it has all three features. Any order outside the test is void.
- Narrow blocking. ISPs block only the feed's domain. Every block is published within 48 hours with the test result, and a platform can appeal to a Designated Review Bench, which decides in 7 days.
- Fast unblock. Re-test within 7 days of a request; unblock within 24 hours of a pass. A missed deadline means automatic unblock.
- No surveillance. No user is identified or punished for watching. The duty is on the platform, and ISPs keep no record of who tried to reach a blocked feed.
- Takedown transparency. Every takedown order is published within 48 hours; an unpublished order is void.
- Fund integrity. The Learning Fund pays only on passed skill checks, each learner counted once per course. Any creator paid more than ₹10 lakh a year gets a faceless audit (Chapter 13).
- Children's data. A breach of children's data protections costs up to ₹200 crore (DPDP Act schedule), set by the Data Protection Board within 90 days.
- Fair competition. No app is removed for any reason other than this chapter's tests without a court finding, a 90-day cure period and an Indian alternative with at least 1 crore monthly users.
Limits to know
A permanent ban for adults is new among major democracies, so the Supreme Court will test it for proportionality, as it did for internet restrictions in Anuradha Bhasin v Union of India (2020): a legitimate aim, a narrow rule and no gentler option that works. The narrow three-part test, the open-content rule and published data on the harm are what carry it. A small share of users will still reach banned feeds by VPN; the revenue fine on the platform, not the ISP filter, does most of the work.
Sources: TRAI 1.09 bn subscribers FY26, Free Press Journal · 1.11 bn June 2026, NewKerala · 588 mn short-video users, IAMAI-Kantar, Manifest Media · Reels usage, Meta · 1.95 lakh blocking orders, OneIndia · DPDP Rules 2025, India Briefing · DPDP Rule 13, data transfers, SFLC.in · ShareChat IPO plan, Business Today · Media and entertainment 2025, EY · Online Gaming Act assent, Business Today · Online Gaming Act takedowns, Organiser · EU finding on TikTok addictive design, European Commission · Douyin 40-minute limit, Sixth Tone · Australia under-16 ban begins, Dawn · Storage of payment system data, RBI · Anuradha Bhasin v Union of India, Columbia Global Freedom of Expression · App blocking playbook, Free Press Journal