The Law of Using AI — India First
Where the law actually stands, dated and sourced: evidence and authentication in the deepfake era, the judiciary's rules and drafts, the profession's duties, data protection, privilege — India first, the world as contrast, and your obligations as a student today.
The hook
“You will misjudge at least one of the six images you are about to see — the room always does. Now imagine the one that fooled you is Exhibit C, and you are the one who must keep it out — or get it in. India has been building the law for exactly this moment: a new certificate regime, a Supreme Court order on who signs it, platform labelling rules in force since February, and a judiciary writing rules for itself. This hour is that law, dated and precise.”
What you'll be able to do
- State what authenticating an electronic record in India now requires — BSA s.63(4) and the Schedule's certificate with its hash value and expert signature, Anvar and Arjun Panditrao's primary/secondary line carried forward as the working assumption, and the Supreme Court's 2026 clarification of who counts as the expert.
- Map the regulatory position, register by register: what the White Paper recommends versus directs, what the draft AI-in-Courts Regulations would require, what the IT Rules now mandate for synthetic media, what the DPDP Act makes you, and what remains genuinely unsettled.
- Apply the professional frame: the Advocates Act and BCI Rules' duties as they bear on AI use, the misconduct line after Pooja Ramesh Singh, and privilege under BSA ss.132–134 as construed in In Re: Summoning Advocates — including the in-house counsel gap.
- State your own position as a student today: the academic-integrity layer, what a defensible record of AI use contains, and how to check any of this session's claims when they move — because they will.
On the syllabus
- Real or AI? — the confidence gap, run on the room, then re-framed as an evidence problem
- Authentication: BSA s.63(4) and the Schedule; Anvar and Arjun Panditrao; the 2026 expert clarification; synthetic media and the liar's dividend
- The platform layer: the IT Rules' synthetic-media labelling and provenance regime, in force February 2026 — and what it does not cover
- The regulatory ladder: White Paper → draft AI-in-Courts Regulations → what is actually notified; the judiciary's own systems
- The profession: Advocates Act and BCI Rules duties; the Pooja misconduct line; DPDP; privilege under ss.132–134 and the in-house gap
- Comparative contrast, and the student's own layer: academic integrity, disclosure, the defensible record
In short
The dedicated legal session, placed here deliberately: by now students use the tools with craft, so the law arrives as the rules of a practice they already have. It opens with the real-or-AI test — the confidence gap made personal — and pivots to evidence: the BSA's certificate regime with its hash value and expert signature, the Anvar–Arjun Panditrao line, the Supreme Court's 2026 order upholding and clarifying the new certificate, and what synthetic media does to all of it, including the liar's dividend and India's new platform labelling rules. Then the regulatory ladder, taught by register: a White Paper that recommends, draft court Regulations that would bind, IT Rules that already do, a data-protection statute phasing in, and the profession's own duties — with Pooja Ramesh Singh as the misconduct line and In Re: Summoning Advocates as the privilege frame. Comparative law appears as contrast: the EU's risk architecture, the US's fragmented judge-by-judge orders, England's cases and warnings. The closing turn is the students' own layer — academic integrity, disclosure norms, the defensible record — and the lab fuses the session's law with the course's craft: statute in the window, an authentication note out.
Why it matters for using AI well
You can answer the questions your seniors are currently googling: what does authenticating a WhatsApp video now take, what may I paste where, must I disclose AI use to this court, what does the misconduct line actually cover, and what changed last quarter. Being the junior who states the register correctly — holding, draft, recommendation — is a professional identity, and it is this hour, retained.
What you can do on Monday
Write the one-paragraph answer to 'can we use this screen-recording in evidence?' for any digital artefact on your phone — what s.63(4) requires, who signs, what you would challenge — and file your own three-line AI-use rule (what I never paste; what I always verify; what I always disclose) where you will see it every day.
What they leave with
The skill
State the Indian position, dated and by register — authentication, platform rules, court rules, professional duty, data protection, privilege — and re-verify it each time before you rely on it.
The insight
Provenance replaced plausibility as the ground of trust, and every register of Indian law is converging on the same demand from different directions: show where it came from, show who checked it, and keep the record that shows you did.
The moment they remember
The confidence-gap reveal turning professional. Image by image the room commits — real or AI, and how sure — and the tally lands: scored like a coin toss, confident like experts. The laugh is at ourselves. Then one slide: 'Exhibit C. You are counsel for the defence.' The laugh stops, and the next forty minutes of statute and doctrine are received as survival equipment — the certificate, the hash, the expert, the label — because everyone in the room has just personally failed the alternative, which was trusting their eyes.
In this session
- 01
The opening experiment re-framed in one slide: the room's real-or-AI score — reliably worse than its confidence — is not a party trick but the factual premise of modern evidence law. If trained eyes cannot tell, authenticity must rest on provenance, not plausibility: where the file came from, who can speak to it, what the hash and metadata show. Every doctrine in this hour is the legal system converting that premise into procedure.
- 02
Authentication, the current Indian frame: electronic records come in under s.63 of the Bharatiya Sakshya Adhiniyam 2023, with the certificate under s.63(4) in the Schedule's form — Part A disclosing the hash value of the record, Part B signed by an expert, alongside the person in charge of the device. The carried-forward case law, taught with its own precision: Anvar (certificate mandatory for secondary electronic evidence, overruling Navjot Sandhu) and Arjun Panditrao (mandatory for secondary; not required where the original itself is produced) — decided under the old s.65B, their carry-over to the BSA the working assumption of practice rather than a settled holding, and flagged as exactly that.
- 03
The 2026 authority the bundle-builders of Session 4 have already met: Pune Bar Association v. Union of India (SC, 22 May 2026) — a three-judge order rejecting a constitutional challenge to s.63(4), holding the hash-value requirement rationally connected to integrity ('an electronic fingerprint'), and clarifying that Part B's 'expert' is not confined to examiners notified under s.79A of the IT Act: qualified cyber-forensic professionals may sign, subject to the court's satisfaction. The order expressly names artificial intelligence and deepfake technology as why the safeguards matter — and it is an order disposing of a petition without notice, not a judgment after contest, which is precisely the kind of distinction this course trains students to state.
- 04
Synthetic media's two-front pressure, named precisely: fabricated evidence that must be kept out, and the liar's dividend — genuine evidence dismissed as fake — that must be defended against; both are provenance problems, and the second is the more corrosive because it taxes every authentic record. India's answer at the platform layer arrived in force on 20 February 2026: the amended IT Rules define 'synthetically generated information', require prominent labels and embedded provenance metadata on it, oblige platforms to prevent unlawful synthetic content, and compress takedown clocks to hours. Two boundaries matter to lawyers: the definition covers audio-visual material, not text — an AI-drafted pleading is not 'SGI' — and a platform label is a new species of provenance evidence whose absence is now itself a signal. No Indian court has yet ruled on the admissibility of a contested deepfake; the personality-rights docket — chatbot impersonation in Sonakshi Sinha, voice cloning in Arijit Singh — is a stream of interim injunctions, remedies rather than evidence law, and the course says so.
- 05
The regulatory ladder for courts, by register — the discipline this course insists on: the White Paper (Nov 2025) recommends and suggests — its firm lines are Guidelines 12 (no confidential or privileged information into any AI tool), 14 (independent verification before reliance), 15 (no AI verifying AI) and 18 (responsibility stays with the human) — and it contains no cloud restriction and no approved-tools regime, whatever summaries claim. The draft Regulations for Use of AI in Courts (June 2026) would bind if notified: fifty-seven regulations, including permitted judicial uses, an absolute bar on AI deciding, risk-scoring or credibility assessment (reg. 20(d) — India's structural answer to the Loomis problem), reg. 43(3)'s requirement that lawyers certify disclosure of AI assistance in prepared material, and reg. 49's mandatory AI training for advocates — a sixteen-hour course like this one is, among other things, reg. 49 arriving early. Publicly opposed in part by the advocates-on-record association, judicially doubted in speeches, and, as of the date on this slide, a draft. Alongside: the judiciary's own stack (SUPACE, SUVAS, TERES, LegRAA, e-filing AI; Adalat AI transcription mandatory in one state's district courts), and the Kerala High Court's 2025 policy for its district judiciary — the first binding Indian AI-use instrument for judges. What is notified changes; the ladder's registers do not.
- 06
The profession's frame, with the pinpoints this course checked: the BCI Rules (Part VI, Chapter II) duties — Rules 3 and 4 on conduct toward the court, Rules 15, 17, 19 and 24 on the client, including the confidentiality rule that still cites the repealed Evidence Act section whose successor is BSA s.132 — enforced through s.35 of the Advocates Act's misconduct jurisdiction; no BCI rule yet mentions technology or competence, which is why Pooja Ramesh Singh matters structurally: it converts 'cite it unverified' into misconduct within the existing rules and directs the BCI to frame AI-specific guiding principles (no committee reported constituted as of the date checked). The BCI's July 2026 circular on digital ethics — reproduced but not readable at the original — reaches students directly: undertakings on deepfake and digital conduct at admission and before internships. And the DPDP Act, dated honestly: enacted, Rules notified, but the substantive obligations commence 13 May 2027 — today the binding grounds for the never-paste rule are privilege and the professional rules; the fiduciary-and-processor structure with its ₹250-crore penalties is the frame you are training for, arriving mid-career-early. Re-checked each cohort.
- 07
Privilege, as construed now: ss.132–134 BSA carry the advocate-client privilege forward; In Re: Summoning Advocates (2025 INSC 1275) holds the privilege is the client's — invocable by the advocate on the client's behalf — sharply limits summoning the accused's advocate, protects the advocate's devices with court-supervised procedures mindful of other clients' confidences, and places in-house counsel outside s.132 altogether, with only s.134's narrower protection. The AI consequences write themselves and the session writes them anyway: a consumer chatbot is nobody's advocate; a paste is a disclosure the client did not authorise; and the in-house gap means a company's legal team enjoys less privilege cover than its external counsel for exactly the material most likely to end up in productivity tools.
- 08
Comparative, as contrast and in three sentences each: the EU AI Act classifies AI for judicial decision-support as high-risk, its obligations phasing in with dates that have themselves moved — the direction is architecture and conformity, not case-by-case duty. The United States has fragmented into judge-by-judge standing orders — over a hundred requiring disclosure or certification, with no national rule; India's draft reg. 43 would be the opposite design: one rule for every court. England and Wales has judicial guidance whose updated text says lawyers need not disclose responsible AI use — the exact opposite of draft reg. 43 — alongside a High Court decision referring fake citations to the regulators and a fresh solicitors'-regulator warning notice; it is the common-law cousin whose case law the Indian Supreme Court itself cited in Pooja. The pattern across all three: verification duties converge; disclosure and institutional design diverge — and a student who can state which jurisdiction requires what, as of when, is ahead of most of the profession.
- 09
The student's own layer, today — and the warning this course owes you: the institutional landscape is real and uneven. One NLU has a published AI-content policy; another's examination regulations make 'use of any means of artificial intelligence' in assessed projects serious academic malpractice; the flagship student journal permits disclosed assistive use but bans generative drafting and AI-supplied citations; the leading international moot's rules are silent, leaving its work-product rule to do the work; and the BCI's digital-ethics push reaches students directly, with undertakings at admission and before internships. So the rule of rules: this course requires AI use under its own assessment scheme — your other subjects are governed by your university's regulations, and in some of them the same conduct is a disciplinary offence. Know your institution's rule before you rely on a tool for assessed work; disclose per the rule, and where none exists, disclose anyway in the form Session 8 drills; keep the prompt log and verification trail as your record; and never let a detection tool's verdict — in either direction — substitute for that record. The bridge to Session 8: the duties are on the table; what remains is the workflow that discharges them.
The mirror
Run it on the class. Then on the machine.
An experiment on the room, the same effect explained in the model, a live demonstration on a real tool, and a named takeaway skill.
Why this shape
The real-or-AI confidence gap is a designed moment that must open this session: the room needs to feel its own inability to spot synthetic media in the minute before it studies the law built to manage exactly that inability. The hour that follows is the course's densest law, taught the way the course teaches everything — every claim dated, every register precise: a holding is not an observation, a draft is not a rule, a recommendation is not a restriction.
Real or AI? — the confidence gap
Experiment on the classOn the class
Six images and clips, one by one: the room commits real-or-AI on their phones, plus a confidence score. The tally lands both numbers side by side — accuracy near chance, confidence nowhere near it.
In the model
Generative models produce artefacts optimised to be indistinguishable — that is the training objective working as designed. Detection tools chase a moving target and mislabel in both directions.
Live on the model
The answers reveal one at a time; then the pivot slide: the best-scoring fake reappears as 'Exhibit C', and the room is asked what, now, could possibly establish it either way — which is the word 'provenance' arriving under its own weight.
The skill
Your eyes are not an authentication method. Demand provenance in proportion to stakes — file, source, hash, chain, label — and budget your scepticism for the genuine article too: the liar's dividend cuts both ways.
The certificate, walked through
DemonstrationThe room predicts
A concrete artefact — a WhatsApp video forwarded to a witness's phone — and the room votes on what its admission now requires: nothing special / a certificate / a certificate plus an expert / it depends on original versus copy.
What is going on
No model in this move — the machine here is the statute: s.63(4), the Schedule's two-part certificate, the hash value, and the primary/secondary line the case law drew.
Shown live
The Schedule's form goes on screen and is filled in against the artefact, field by field — and the room watches the 2026 clarification matter in practice: who, in this fact pattern, can sign Part B?
The skill
Read the form, not the folklore: hash in Part A, expert in Part B, and the Anvar–Arjun Panditrao line deciding whether the certificate is needed at all.
The register drill — holding, draft, recommendation
DemonstrationThe room predicts
Five one-line claims about Indian AI law go up ('the Supreme Court has banned…', 'lawyers must now disclose…', 'the White Paper prohibits…'), and the room sorts each on their phones: holding / interim order / draft / recommendation / false.
What is going on
This is the Ladder applied to legal-regulatory claims — the register of an instrument is exactly the kind of nuance that summaries, human and machine, flatten first.
Shown live
Each claim resolves on screen against its primary source, register named and dated — including at least one claim that a chatbot, asked live, states in the wrong register with full confidence.
The skill
Never repeat a rule without its register and date. 'Draft reg. 43 would require' and 'reg. 43 requires' are different sentences, and the difference is your credibility.
The legal thread
This session is the thread made whole: the confidentiality rule of Sessions 3–4 lands on DPDP and privilege doctrine; the verification habit of Session 6 lands on Guideline 14 and the misconduct line; the disclosure instinct lands on draft reg. 43 and academic-integrity norms. One session carries the treatment; the rest of the course made it consequence rather than compliance.
Hands-on · 30 minutes · on your own laptop
The Authentication Note
The session's law and the course's craft, fused: put the actual statutory text and the operative case paragraphs in the model's window, and draft the five-line authentication note a supervising advocate would actually want — what admitting this artefact requires, who signs what, and what your challenge would be from the other side. Context engineering, applied to a provision; the deliverable is the exam-grade artefact of this course.
1 · Watch — the instructor demonstrates
On
Gemini or ChatGPT (free tier, temporary chat); the supplied text pack: BSA s.63 with the Schedule, and the operative paragraphs of Anvar, Arjun Panditrao and Pune Bar Association
The exact prompt
The text between the markers contains s.63 of the Bharatiya Sakshya Adhiniyam 2023 with its Schedule, and operative extracts of three Supreme Court decisions. Using only this text: for the artefact described below, state (1) whether a s.63(4) certificate is required, and why, citing the provision or paragraph; (2) exactly what the certificate must contain and who must sign each part; (3) the two strongest objections opposing counsel could raise to the artefact's authenticity. Artefact: a video received on WhatsApp, downloaded to counsel's laptop, to be tendered from that laptop; the original sender's phone is unavailable. If the supplied text does not resolve a point, say so rather than answering from memory. === TEXT BEGINS === [pack] === TEXT ENDS ===
Point at
Point at three things: whether it correctly reaches 'secondary — certificate required' on these facts; whether the certificate answer includes both Part A's hash and Part B's expert (the detail most human summaries drop); and how it handles the sender's-phone problem — the honest answer flags what the pack cannot resolve.
Roughly what comes back
A sound skeleton note citing the provision and paragraphs, usually correct on the certificate's contents, and — usefully for the reveal — often glossing the expert requirement or overstating how settled the BSA carry-over is. Its errors are the teaching material; the students' job is to catch them against the pack.
If it misbehaves
The pre-captured demo note (fallback folder, S7). The text pack is the fallback for everything else: if all AI tools are down, the lab runs entirely on the pack and paper — statute-reading needs no tokens.
2 · Your turn — a variant, not a copy
Your turn, on a different artefact — draw one of four from the deck: a CCTV clip exported by a mall's security contractor from a DVR that has since been wiped; a voice note in a family WhatsApp group tendered in a maintenance dispute; a screen-recording of a vanished Instagram story; a video downloaded from a platform, carrying the platform's AI-content label. Same pack, same method: statute in the window, pinpoints demanded, and then your own five-line authentication note — in your words, checked against the pack, with the model's gloss corrected where you caught it.
Free tier
One pack-paste and three or four prompts — inside every free window (the pack is sized to fit, which is why it is extracts, not full judgments). No accounts beyond the usual; under-18 students on Gemini/ChatGPT as always. If rate-limited, the note is drafted from the pack directly — the statute does not need a subscription.
3 · The reveal
Artefact by artefact, the room's key moves tally on phones: who required the expert signature, who spotted the primary/secondary turn, and — on the labelled-video artefact — who used the IT-Rules label as provenance evidence rather than as proof of fakery. One strong note and one instructive miss go on screen; the miss is always the certificate treated as a formality rather than a fact to be established.
Deliverable
A five-line authentication note for your artefact — requirement, contents, signatories, challenge, unresolved point — with the corrected model transcript attached. The evaluation session's applied question is this artefact class; this note is your revision material.
Run of show · 30 minutes
- 0–10 min — Watch: the demo artefact through the statute-in-the-window method; the model's note read against the pack, its gloss caught and corrected.
- 10–13 min — Your turn: draw your artefact, open a temporary chat, load the pack.
- 13–21 min — Run the method: certificate question, contents, signatories, objections — pinpoints demanded, 'not in the pack' respected.
- 21–27 min — Draft your own five-line note; correct the model against the pack; note one point the pack could not resolve.
- 27–30 min — Reveal: key-move tallies by artefact; one note and one miss on screen.
For the instructor · before the session
- Assemble and verify the text pack: s.63 + Schedule against the Gazette-checked text, and the operative extracts of the three cases against the reports; the pack is a teaching document — it goes in the codebook.
- Curate the real-or-AI set fresh each cohort (last year's fakes are this year's obvious); pre-test the quiz link and keep local copies of every item.
- Check the regulatory ladder's status the week of the session: draft Regulations notified? BCI committee reported? DPDP phase-in reached? The register drill's answers are re-dated each run.
- Pre-run the demo prompt; capture the note and its characteristic gloss for the fallback.
- Print the four artefact cards (or post them); each carries its fact pattern in three lines.
- Launch the Session 7 poll deck: the real-or-AI commits, the certificate vote, the register drill, and the lab tallies.
Key sources & cases
Bharatiya Sakshya Adhiniyam, 2023, s.63 and the Schedule; Anvar P.V. (2014) 10 SCC 473; Arjun Panditrao (2020) 7 SCC 1
The authentication frame: s.63(4)(c)'s certificate signed by the person in charge and an expert, in the Schedule's form (Part A hash value, Part B expert); Anvar's mandatory-certificate rule for secondary evidence; Arjun Panditrao's primary/secondary line. Sections and cases verified 2026-08-26 (codebook); carry-over to the BSA taught as the working assumption, not a holding — commentary contests parts of it.
Pune Bar Association v. Union of India, W.P. (C) No. 599/2026 (SC, 22 May 2026, Surya Kant CJI, Bagchi and Pancholi JJ.)
Challenge to s.63(4) rejected: hash value rationally connected to integrity; Part B's 'expert' not confined to s.79A-notified examiners; AI and deepfake technology expressly named as what accentuates the authenticity problem. Read to the signed order 2026-08-27. An order without notice, not a contested judgment — teach the register.
IT (Intermediary Guidelines and Digital Media Ethics Code) Amendment Rules, 2026 — G.S.R. 120(E), 10 Feb 2026; in force 20 Feb 2026
India's synthetic-media regime: 'synthetically generated information' defined (audio-visual; text excluded; good-faith drafting and research carve-outs); prominent labels and embedded provenance metadata required; platforms must prevent unlawful SGI; takedown clocks compressed to hours. Text read from MeitY's consolidated Rules 2026-08-27; the commencement date rests on MeitY's FAQ — confirm the Gazette line before academic-council use.
Supreme Court of India: White Paper on AI and Judiciary (Nov 2025); draft Regulations for Use of AI in Courts, 2026 (June 2026)
The ladder's two upper rungs, by register: the White Paper recommends (Guidelines 12/14/15/18 quotable; no cloud restriction, no approved-tools regime — both misattributions are corrected in this course's codebook); the draft Regulations would bind — permitted uses, decision and risk-scoring bars, reg. 43 party disclosure — and were a contested draft as of 2026-08-27. RE-CHECK notification status each cohort; this slide is rewritten the day it changes.
Pooja Ramesh Singh, 2026 INSC 668; Advocates Act 1961, s.35 with BCI Rules Part VI Ch. II; BCI circular of 17 July 2026
The professional-duty stack: the duties (court, client, candour, confidentiality) under the Rules; s.35's misconduct jurisdiction as the enforcement path; Pooja's holding placing unverified AI citation inside it and directing BCI guiding principles (committee status re-checked each cohort); the BCI's digital-ethics circular known from reproductions — the original was unreadable at checking; VERIFY before quoting its text.
Digital Personal Data Protection Act, 2023 + DPDP Rules, 2025 (G.S.R. 846(E), 13 Nov 2025)
The data-protection frame for AI use, dated honestly: fiduciary/processor structure, breach duties, penalties to ₹250 crore — with substantive obligations commencing 13 May 2027, so taught as the incoming frame rather than today's binding ground (today's grounds: BSA s.132 privilege and the BCI Rules). Rules' number and phase-in read from MeitY/PIB documents 2026-08-27; section pinpoints for the lawyer-specific reading flagged VERIFY TO SOURCE against the Act's text.
In Re: Summoning Advocates, 2025 INSC 1275 (SC, 31 Oct 2025); BSA ss.132–134
Privilege as construed: the client's privilege, invocable by the advocate; summons to defence counsel confined to s.132's exceptions with senior-officer satisfaction and judicial review; device productions court-supervised with other clients' confidences protected; in-house counsel outside s.132, s.134 only. Verified 2026-08-26 against the official PDF; now also reported in SCC — add the parallel citation at teaching time.
Comparative set: EU AI Act (Reg. (EU) 2024/1689), Annex III pt. 8; US judge standing orders (trackers, 2026); England & Wales — judicial guidance (2025), Ayinde [2025] EWHC 1383 (Admin), SRA warning notice (Aug 2026); Cork v. Smith [2026] EWHC 1199 (Ch)
Contrast only, three sentences each in class: EU architecture with phased (and shifting) dates; US fragmentation — tracker counts, not official figures, cite as such; England's guidance-plus-referral approach, whose Cork decision the Indian Supreme Court itself cites in Pooja. ABA Formal Opinion 512 could not be verified at checking and is not taught until it is. Every comparative claim carries its date.
MIT Media Lab Detect Fakes (the real-or-AI test); Chesney & Citron, 'Deep Fakes' 107 Cal. L. Rev. 1753 (2019)
The opening experiment's instrument, and the liar's dividend's source — attribute the term to Chesney & Citron. Both carried over from the earlier design; verification status per codebook (capture exists; confirm the quiz's current URL at teaching time).
Readings
- BSA 2023, s.63 with the Schedule — the section and the form; twenty minutes, and the spine of the session
- Arjun Panditrao (2020) 7 SCC 1 — the primary/secondary paragraphs (re-read from Session 4's bundle)
- Pune Bar Association v. UOI (SC, 22 May 2026) — the order, seven pages
- In Re: Summoning Advocates, 2025 INSC 1275 — the directions in para 67, and the in-house counsel passage
- White Paper (Nov 2025) — Guidelines 12–18; draft AI-in-Courts Regulations (June 2026) — reg. 43 and the definitions; Chesney & Citron (2019) — the liar's-dividend section
Sixteen hours, one professional discipline.
Using AI well is not a knack — it is craft, competence and verification, practised until they are habits you could defend in court.