Students in India are often told two opposite things. One is that using AI is cheating. The other is that they will not get a job unless they can use it. Both can be true, and neither one is a complete instruction.
This week, exam authorities in England and South Korea tightened the case for punishment. A philosophy department in the United States banned even brainstorming with AI. At the same time, some Indian institutes, including IISER Pune, IIM Ahmedabad, IIM Bangalore and the Indian School of Business, are writing course-by-course rules and asking students to disclose what they used.
My take is simple. Punishing students for using AI will not make them think better. It sends them into jobs that expect AI fluency without ever having practised it honestly. India should copy the disclosure model, not the ban.
What changed this week
The Guardian reported on 18 November 2026 that Sir Ian Bauckham, Chief Regulator of Ofqual, England's qualifications regulator, warned that students could face "life-altering" penalties if they were caught relying on AI or smart glasses to cheat. The paper said he told the Press Association: "If you are caught committing malpractice with an important, high-stakes qualification … in extreme cases you can lose the entire qualification which, as I am sure you can imagine, can be life-altering for somebody's future."
Ofqual is launching a publicity campaign, as coursework deadlines approach, to make clear that using AI for assessments such as extended essays and failing to declare it will not be tolerated. Bauckham is also writing to examination boards in England to say that invigilators should be trained to spot wearable devices in next summer's exams, and he called it "completely unacceptable" for any retailer "to market items explicitly for the purpose of exam cheating."
Those warnings sit on top of official numbers, not rumours. Ofqual's official summer 2025 malpractice statistics, published on 11 December 2025, recorded 2,225 student cases involving mobile phones and other communication devices, the most common offence type. The same official tables recorded 100 proven student cases of plagiarism involving the misuse of AI in summer 2025, up from 85 in summer 2024. Those AI cases were 2.0% of all student malpractice cases in 2025, and 1.7% in 2024. Hidden AI use is a real integrity problem. It is also not the same as most of the cheating Ofqual already records.
On the same day, Yonhap News Agency reported from Seoul that South Korea's government and the ruling Democratic Party had agreed to seek legislation to extend the ban on taking the college entrance exam, the CSAT, known as Suneung, for AI-assisted cheating, to a maximum of 10 years, "with the aim of applying it retroactively from this week's test." The meeting was held a day before the exam, which was due on Thursday, 19 November 2026. Party spokesperson Park Sung-joon said: "We have agreed to seek a revision to the Higher Education Act to apply the penalties retroactively to this year's exam." Yonhap also reported plans for an additional 8,400 personnel at test sites, and tighter checks for smart glasses, microcameras and other devices. This is a plan to change the law. It is not, on its own, a penalty already in force.
Two days earlier, on 16 November 2026, the philosophy site Daily Nous reported that the Department of Philosophy at Trinity College in Hartford, Connecticut, had adopted a policy banning deliberate AI use in its courses. The policy, as published there, says: "Deliberate use of AI in philosophy courses, in any capacity and for any purpose, is forbidden." That includes summarising readings, brainstorming, research for presentations or oral exams, and help with writing or editing, including compiling a bibliography. "Philosophy is focused on cultivating our abilities to think, to read, to write, to speak, to reason, and to live ethically. AI undermines all of these goals." Daily Nous, citing the student newspaper the Trinity Tripod, said the faculty vote was unanimous. Professor Shane Ewegen told the Tripod, as quoted there, that the department wanted to avoid "having six different AI policies that contradict each other," and agreed on "a single blanket policy, which simply forbids the use of AI in any capacity in any philosophy course."
I take the worry seriously. Thinking is the point of an education. A blanket ban, though, does not teach a student how to use a tool they will meet at work, or how to be honest about it.
What some Indian institutes are doing instead
India already has a clearer alternative than "ban everything" or "use anything."
IISER Pune's official Guidelines for Generative AI usage, adopted by the Senate on 11 March 2026, start from permission, not from a ban. "Unless explicitly forbidden by the competent authority, the use of GAI is permitted in all documented outputs and other activities of the students and staff of the institute," the guidelines say. If a teacher or supervisor does not want AI used, they must say so "unambiguously, in writing." Without that written prohibition, "punitive actions cannot be taken against anyone for the mere act of using GAI." Academic integrity still applies. The student remains responsible for accuracy, and must be able to understand the work. The same document says that if a student uses AI on an assignment and then cannot explain the concepts in an oral examination, the responsibility sits with the student.
The guidelines also draw a practical line on disclosure. Using AI to search literature, understand a concept, brainstorm, or comment on text you have already written does not, on its own, have to be attributed. When AI output forms a substantial part of the final work, for example complete sentences or paragraphs, substantial code, or figures, attribution is mandatory. Professor Sutirth Dey, chairperson of IISER Pune's AI Committee, told The Indian Express in October 2026: "Some faculty members explicitly instruct their students about where they should use AI and where they should not." The same report said most quizzes and mid-semester and end-semester examinations at the institute still run without AI, and described a bioinformatics course in which students could use AI to write programmes but then had to solve similar problems without it in the exam.
That is disclosure and course design, not a slogan.
IIM Ahmedabad's director, Professor Bharat Bhasker, made the same choice in an Indian Express interview published on 15 October 2026. "We have responded by encouraging students to use AI rather than prohibiting it," he said. "We purchased around 1,500 licences, providing access to students, faculty members, research associates, and others at the institute. The initiative was introduced in April this year." Students may use tools such as ChatGPT or Gemini to summarise cases before class. Faculty then raise the discussion by changing assumptions, so the work is judgement, not a pasted summary.
The Indian Express also reported that ISB has not set a uniform, institution-wide restriction. Individual faculty members decide whether AI is permitted, restricted or prohibited in a given course or assignment. Professor Vishal Karungulam, academic director at ISB's Centre for Learning and Teaching Excellence, said the approach was meant to prepare students for workplaces where generative AI is increasingly relevant, while still developing responsible use and critical judgement.
The Financial Express reported on 4 October 2026 that, during the June to July term, 78 MBA students at IIM Bangalore, across core Business Analytics and general MBA elective cohorts, took their Generative AI midterm by building and deploying their own chatbots to answer the questions. Professor Subhabrata Majumdar built a password-protected platform, and the model was restricted to assigned course materials. This was a designed exam in one set of courses, not a claim that every paper at IIM Bangalore now runs this way. It is still a useful contrast with a total ban: the students had to practise using the tool, under rules, in the open.
Is using AI cheating?
Sometimes yes. If the exam hall forbids devices, a phone or a pair of smart glasses is cheating. If the assignment says you must write the essay yourself, and you submit an AI draft as your own, that is cheating. If you cannot explain the work you turned in, you did not do the learning, whether or not a detector flags it.
Sometimes no. Using AI to quiz yourself after you have tried a problem, to explain a step you missed, or to comment on a paragraph you wrote, is closer to a tutor than to a ghostwriter. IISER Pune's official guidelines treat several of those uses as allowed, and as not always requiring a formal attribution, so long as you still check the output and do not pass off substantial AI text as your own.
The line is not "did a tool touch this." The line is: what did this course allow, did you hide the use, and can you still think the work through yourself? I made a similar point, in a different debate, in Should colleges ban AI? What actually helps students learn. A ban does not reach the phone in a student's bag. Disclosure, and a better test of understanding, can.
Five checks before you submit
- Read the rule for this assignment, not a general vibe.
If your teacher has put a written AI rule on the course page, that is the rule. If they have not, do not assume every use is safe, and do not assume every use is banned. Ask. IISER Pune's official policy is unusual in saying that, without a written prohibition, a student should not be punished for the mere act of using AI. Most schools and colleges in India have not published a rule that clear. Until they do, get the instruction in writing.
- If you use AI, do not hide it.
When AI has written substantial sentences, code or figures that end up in the work you submit, say so. Name the tool and what you changed. Hidden use is what Ofqual is targeting in coursework. Honest disclosure is what the IISER Pune guidelines require once the output is substantial.
- Use AI as a tutor, not as a stand-in for your thinking.
Try the problem first. Then ask the tool to quiz you, explain a step, or comment on your draft. I set out that habit in more detail in how to use ChatGPT without switching off your brain. If the tool did the thinking, you will feel it the moment a teacher asks a follow-up question.
- Check the work, then see whether you can defend it.
AI invents sources and sounds confident when it is wrong. The IISER Pune guidelines put the duty to verify on the student. Before you submit, close the laptop and explain the answer out loud. If you cannot, you are not ready, and a job that expects AI fluency will not be kinder than a viva.
- In an exam hall, follow the hall rules.
A high-stakes written exam is not the place to test a grey area. Leave the phone outside. Do not wear a device the rules forbid. That is true in England, in South Korea, and in an Indian board or university exam. Punishment is the right tool for concealed devices. It is the wrong tool for the whole of learning.
If you are a parent, ask the school or college for the written AI policy, and ask how teachers will check understanding, not only how they will punish misuse. A short viva, an in-class task, or a request for drafts will tell you more than a detector score. Teachers who want practical classroom ideas can start with AI for teachers and colleges.
Copy the disclosure model, not the ban
I do not want students to outsource their thinking. I also do not want India to import a model that treats every honest use of AI as a scandal, then sends those same students into offices that expect them to use the tools well.
Ofqual is right that undeclared AI in a high-stakes qualification is malpractice. South Korea is right that hidden devices in a national exam are an attack on fairness. Trinity's philosophy faculty are right that brainstorming, reading and writing are skills you have to practise. The useful Indian answer is the one IISER Pune wrote down: say whether AI is allowed, say how to disclose it, test whether the student understood the work, and keep the responsibility with the person whose name is on the page.
At The Arthi AI Collective, I help students and working professionals use AI as a thinking partner, and leave with work they can explain, not only a file they cannot defend. If that is what you are looking for, explore the programs.