The AI Tutor Bargain: What India Risks to Give Every Child a Free Coach

Why This Proposal Cuts Deeper Than It Sounds

An op-ed calling for free AI tutors reads, at first glance, like routine tech-optimism. It is not. Buried inside Srivatsa Krishna's "public rail, private engine" model is a genuine governance dilemma that UPSC aspirants need to sit with rather than skim past: can the state build the infrastructure that decides which 17-20 million coaching-dependent students get taught what, without quietly becoming the gatekeeper of their opportunity?

The Angle Worth Understanding

The proposal wants government to do exactly what it did with UPI - supply identity (Aadhaar/DigiLocker), a discovery layer (APAAR ID, DIKSHA) and a trust layer (open ratings on provider outcomes)

• while leaving content and mentoring to competing private players. On paper this "unbundles" a Rs 1,23,000-crore-plus coaching industry that currently bundles content, doubt-solving, peer discipline and a credibility signal into one expensive package families cannot easily unpick.

What is genuinely gained: A weak-in-organic-chemistry student in a Common Service Centre could, in principle, be routed by an open recommendation engine to the best-rated module on that exact topic - regardless of which company made it. That is a real democratisation of the "last child in line," not a slogan.

What is genuinely risked: The same rails that route a student to content also decide, through an "open recommendation/diagnostic engine," what that student sees next and the same government body that sets taxonomy and accreditation standards decides which providers get delisted.

A public infrastructure that owns discovery, credentialing and data portability for tens of millions of students is not a neutral pipe - it is a chokepoint. Aadhaar-linked sign-up and a "learning wallet" tied to APAAR mean the state, not just the market, ends up holding a permanent, granular record of a child's academic weaknesses.

Why This Is Not a False Choice

The easy line - "government should just enable, not control"

• dodges the actual mechanism. The Section 8 not-for-profit steward proposed here would set the syllabus taxonomy, run accreditation with "delisting teeth," and own the reference app. Every one of those functions is also a lever of control. Unbundling the coaching cartel does not automatically prevent the state from becoming a new, less visible one.

The Honest Resolution

The model is worth building, but only if the specific safeguards Krishna names - student-owned data under a DEPA/Account Aggregator consent pattern, protocol-level portability so no provider or ministry can lock in a student's learning history and open outcome data rather than closed ministry dashboards - are treated as non-negotiable architecture, not add-ons.

Strip those out and "public rail, private engine" quietly becomes a single national algorithm deciding a generation's academic trajectory. The democratisation is real only if the exit door - a student's right to take their data and leave the rail entirely - is built in from day one, not promised for later.