AI and Academic Integrity Policy Drafting

A dean asked for 'the UGC ban on ChatGPT' to annex to a show-cause. We asked for the URL. There wasn't one. What exists, and has existed since 2018, is the University Grants Commission (Promotion of Academic Integrity and Prevention of Plagiarism in Higher Educational Institutions) Regulations, 2018 — similarity tiers, departmental and institutional academic integrity panels, and duties to educate. Those regulations were written for copied text. They do not, by themselves, define unacknowledged generative output. They also do not vanish because a model exists.

This is a drafting template for academic councils, integrity panels and controllers. It is not a statute. It is not legal advice. It will not survive if you paste it without reading your ordinances, your professional-council rules, and the live UGC page. As of 17 August 2026 we have not found a replacement of the 2018 regulations that we will cite as a ChatGPT ban. Secondary blogs that announce 'UGC AI guidelines 2026' are not the Gazette. If UGC has issued a later letter that applies to you, pin that letter. Do not let a coaching-centre blog become your law.

The institutional job is therefore ordinary: say what students and faculty may do, what they must disclose, how you will assess, and how the 2018 machinery still hears a case. Inventing a ban you cannot show is how you lose the first appeal.

What the 2018 regulations do — and what they leave you

They require HEIs to promote integrity, to use similarity tools as the institution arranges, and to run DAIP/IAIP processes with penalties scaled to similarity bands for students and staff in the contexts the regulations name. Read the text. Do not quote a workshop paraphrase of the percentages. The official PDF is on UGC's site.

They do not mention ChatGPT, Gemini, or 'AI detectors'. A high score on an AI-detector product is not a 2018 similarity finding. If you want detector scores in a hearing, you must write that into institutional policy, and you must be ready to defend error rates. We advise caution. Detectors are uneven across Indic English and neurodivergent prose.

Unacknowledged verbatim use of generated text can still be approached as plagiarism or as a false statement of authorship under your ordinances, if you have said so. The 'if you have said so' is the work. Silence is how every case becomes a debate about novelty.

Draft clauses, not slogans. Each clause needs an owner and a hearing path.

Clause	Say	Do not say
Scope	Assessed work, theses, publications under the HEI's name	Any use of a spellchecker is misconduct
Authorship	The student or teacher is the author and is responsible for every sentence submitted	The model is a co-author
Disclosure	When the assessment allows assistance, disclose the tool and the kind of help	A blanket confession box that is never read
Prohibited	Undisclosed generation of analysed argument, data, or citations; invented references	'AI is banned' without defining the verb
Detection	Similarity tools under 2018; human academic judgement; student viva where needed	An AI-detector number as conclusive proof
Hearing	Existing DAIP/IAIP or ordinance process, with the evidence shown	A department WhatsApp poll

Assessment design is the real policy

A take-home essay that can be wholly generated, with no viva, no process folio, and no local data, is an assessment design failure. Policy will not save it. Change some tasks: oral, in-class, staged drafts, locally observed data, reflection on retrieval from the library agent you actually run.

Different programmes need different clauses. A coding lab, a studio, and a theory MA are not one moral universe. Allow the board of studies to tighten, not to quietly ban electricity.

Faculty use is in scope. A teacher who generates a reference list that does not exist has a 2018-shaped problem and a professional one. A teacher who uses a campus-approved agent to retrieve the ordinance is doing their job.

Paste-ready skeleton — still not your law until you adopt it

1. Purpose. This policy implements the HEI's duty under the UGC 2018 integrity regulations and the ordinances, for a world in which generative tools exist. It does not replace those instruments.

2. Definitions. Generative assistance means a system producing new wording, code, data, images or citations. Recoverable assistance means spelling, grammar, and library retrieval that does not invent a source. The academic council may refine these lists.

3. Default rule. Recoverable assistance is allowed. Generative assistance is allowed only when the assessment brief says so, and must be disclosed in the form the brief specifies. Undisclosed generative assistance, invented sources, and submission of generated work as solely one's own are misconduct.

4. Process. Allegations go through the existing integrity panels. Evidence includes the brief, the disclosure, similarity reports, and any viva. Detector scores are not conclusive. Students see the evidence.

5. Education. Induction will teach this policy. Library and writing desks will teach citation of tools. Ignorance after induction is not a defence; induction still has to happen.

6. Review. Annual, or when UGC issues a live instrument that actually names generative systems. The cell will not treat a news item as that instrument.

What not to annex

Do not annex a vendor's acceptable-use policy as if it were UGC. Do not annex an American university's honour code without reading it — they sit on different statutes. Do not annex an AI-detector marketing PDF.

Do annex: the 2018 regulations, your ordinances, any professional-council exam instruction, and the live UGC letter if one exists on the day you adopt.

Briefs are the policy students actually read

The six-clause policy is for the panel. The assessment brief is for 11 p.m. Put three words at the top of every brief this term: allow, disclose, or forbid — and say what those words attach to. 'Disclose grammar help. Forbid generated arguments and invented citations.' That is a rule a first-generation student can obey. A 14-page PDF is not.

Programme boards may tighten, not loosen below the default, unless the academic council said they may. A coding lab that allows an approved on-prem pair-programmer with a disclosure is a tightening of method, not a holiday. A theory paper that says nothing is the default, not a secret ban.

Keep student submissions off vendor training corpora. That is a DPDP and a processor-contract sentence as much as an integrity sentence. The library agent and the integrity policy meet here: retrieval of a real source is not generation of a fake one.

Objections you will hear — and what to do with them

These are the lines that stall the file. Answer them in the room, then put the answer in the note. A spoken answer without paper will be forgotten by the next officer.

If we do not ban it, we are soft.

A ban you cannot define is soft in court. A disclosure rule plus assessment redesign is hard in the only place that matters: the work.

UGC will notify something next month. Let us wait.

Wait for that instrument when it exists. Meanwhile students are submitting. Adopt a living institutional policy that yields to a later UGC regulation automatically.

We will fail NAAC if we allow AI.

We have not found a NAAC criterion that says that. Inventing a NAAC fear is not a policy. Write integrity practice you can show a peer team.

Faculty will not read another policy.

Put the default rule on the assessment brief: allowed / disclose / forbidden. The long policy is for the panel. The brief is for the night before the deadline.

A six-week adopt-and-teach cycle

Academic council, not the computer centre, owns this policy. The centre can host the tools.

1. Week 1: pin the 2018 PDF and any later live UGC letter. Kill blog citations in draft notices.
2. Week 2: draft the six clauses. Boards of studies add programme notes.
3. Week 3: integrity panel walks two mock cases — invented citations, and disclosed grammar help.
4. Week 4: academic council adopts. Briefs for the current term get a three-word stamp: allow / disclose / forbid.
5. Week 5–6: induction for students and faculty. Library teaches citation of tools. No detector gate.

How this shows up in the file

Resolution: we adopt an institutional generative-assistance policy under the 2018 regulations; we do not claim a statutory ChatGPT ban; detector scores are not conclusive; briefs carry the three-word stamp; the policy yields to a later UGC instrument.

Attach the 2018 PDF hash and the date you downloaded it.

What the next file must contain

“AI and Academic Integrity Policy Drafting” earns a line in the noting only if a P3 University can attach proof of “academic integrity AI policy.” A heading is not proof. A vendor slide is not proof. A workshop photograph is not proof.

Write three dated sentences: what was decided, who owns it after the next posting order, and when it will be re-checked. If you cannot write the three sentences, you are not ready to buy, to sell, or to go live.

Leave unsourced percentages out of the note. DPDP is not a blanket localisation statute. The November 2025 AI governance text is guidance, not an Act. CERT-In’s 28 April 2022 directions still set specified incident and log clocks. A PAC, when lawful, lives in GFR Rule 166.

• Name the designation that owns “academic integrity AI policy.”
• Attach one artefact a stranger can open next year.
• Record the instrument you are actually using.
• Revisit when the model, the SI, the notice or the posting changes.

This article is informational field guidance for Indian public universities and colleges, not legal, regulatory, academic, labour or procurement advice. Confirm against the live AICTE / UGC / NAAC notice, the DPDP Act and Rules as commenced, your Act and statutes, reservation orders, labour counsel and the syndicate before you file it.