Universities in the age of AI: The tension between trust, competence and surveillance

Illustration: TBS

For 133 years, Princeton University upheld an extraordinary tradition. There were no proctors in examination halls. Professors trusted that students, guided by their own conscience and integrity, would not cheat. At the end of each exam, students signed a pledge affirming that they had not engaged in any dishonest practices.

Since 1893, this "Honor Code" was not merely an academic regulation; it was a cultural commitment to trusting people. World wars came and went, the internet arrived, smartphones emerged — yet the tradition endured. In the end, however, the force that transformed this practice was AI.

In the era of ChatGPT and other generative AI tools, secretly generating answers during exams has become so easy that a mere "pledge of honesty" no longer seems sufficient. Technology has reached a point where the boundary between what a human writes and what a machine produces can become blurred in an instant.

Facing this reality, on 11 May 2026, Princeton faculty voted in favor of bringing proctors back into examination halls. In other words, teachers or invigilators will now be present during exams to supervise and monitor students. Technically, the "Honor Code" still exists, and students will continue to sign the pledge, but the old culture of trust-based examinations is no longer what it once was.

To us, this incident stands as one of the most profound symbols of the present AI era. We often talk about the efficiency, speed, and convenience of AI, but we rarely discuss the social relationships this technology is gradually weakening. Princeton's decision reminds us that technological change does not merely introduce new conveniences; sometimes, it also rewrites the very structure of trust between people.

At one time, the university believed that if students were given freedom, they would act responsibly. Now, the institution feels that in this technological age, relying solely on morality is no longer enough. Perhaps this is the greatest transformation of the AI era: we are not only becoming dependent on machines, but are also gradually beginning to lose trust in one another.

The incident at Princeton is not an isolated case. Similar changes are now emerging at universities around the world. Discussions are also underway at Oberlin College about revising its traditional Honor Code. For years, the presence of teachers in examination halls was almost prohibited there, because the institution believed that students' integrity was the foundation of the examination system.

However, as cases of AI-assisted cheating have increased, the college is now considering allowing proctoring and direct supervision during exams. In other words, universities that once viewed freedom and trust as an essential part of education are gradually moving toward systems based more on surveillance and monitoring.

At the same time, the nature of examinations itself is also changing. Many universities are now returning to oral examinations or "viva"-style assessments. At the University of Pennsylvania, many instructors are adding oral exams alongside written assignments to determine whether students truly understand the subject or have simply used AI to generate polished answers.

In some courses at New York University, AI-driven oral examinations are also being introduced, where chatbots question students and assess the depth of their thinking. The situation creates a certain irony: AI itself is being used to solve problems created by AI.

In this age of digital technology, many institutions are once again returning to older methods. Handwritten examinations, answering on paper in physical classrooms, and in-person assessments are regaining importance. Universities are realising that it is becoming increasingly difficult to distinguish between a perfectly polished assignment completed at home and genuine learning.

If the education system focuses only on increasing surveillance, it may create fear, but it will not foster genuine AI literacy. What Bangladesh needs now is an educational policy where students do not simply learn how to generate answers using AI, but also understand its limitations, errors, biases, and broader social impacts. Otherwise, we risk producing a generation that can use technology effectively, but cannot question it.

At the same time, institutions such as the University of Florida are also revising their policies. Many courses now explicitly specify where AI tools may be used and where they are prohibited. Meanwhile, several universities are formally acknowledging that relying solely on AI-detection software to accuse students is risky, since these technologies often produce inaccurate results.

At the University of Illinois Urbana-Champaign, where I (Sharifa) work, a kind of "hybrid approach" is now emerging to address the impact of AI. Assessment methods are being adjusted according to the realities of each discipline. In STEM subjects, greater emphasis is being placed on hands-on examinations, restricted coding environments, explaining programming logic on the board, and problem sets based on new hypothetical datasets, making it more difficult to directly copy answers from AI tools.

On the other hand, departments in the humanities and social sciences are increasingly adopting in-class writing, draft-based assessments, and discussion-oriented evaluations. At the same time, universities are now explicitly stating in their syllabi where AI tools may be used, where they are prohibited, and whether attribution or disclosure is mandatory when AI is used.

I (Ishtiaque) teach at the University of Toronto where I'm witnessing similar changes. The university's 2025 AI Task Force report recommends separate assessment strategies for STEM and non-STEM disciplines. Most importantly, universities are gradually being forced to move away from evaluating only the "final answer" and return instead to assessing the "process of thinking" itself.

The most important question here is not about technology, but about the philosophy of education itself. Are we moving toward a world where every student is treated from the outset as a "potential cheater"? Surveillance may reduce certain forms of academic dishonesty, but at the same time, it risks making educational environments more fearful, competitive, and rooted in distrust.

Universities once sought to cultivate independent thinking, morality, and personal responsibility among students. Now, it seems that space is gradually being overtaken by surveillance, verification, and suspicion. Perhaps the AI era has not only introduced new technologies; it has also taught us that modern institutions are increasingly beginning to speak the language of control rather than the language of trust.

These changes could create an even more complex reality for Bangladesh's education system. Much of our educational structure still remains heavily dependent on memorisation, examinations, and grades. In such a context, AI risks rapidly transforming from a "learning assistant" into a "shortcut machine." Already, many students are using AI to write assignments, generate code, and even draft sections of research papers, yet most institutions still lack clear guidelines on where AI use is acceptable and where it constitutes academic dishonesty.

As a result, while students are quickly learning how to use AI, universities and schools are struggling to adapt their policies, teacher training, and assessment systems at the same pace. This growing gap could eventually create a major crisis of trust within the education system.

At the same time, it is also true that Bangladesh can no longer afford to stay distant from AI. It is set to play a crucial role in future jobs, research, business, and even everyday communication. So the question is not whether AI should be banned, but rather how students can be taught to use it responsibly, critically, and ethically.

If the education system focuses only on increasing surveillance, it may create fear, but it will not foster genuine AI literacy. What Bangladesh needs now is an educational policy where students do not simply learn how to generate answers using AI, but also understand its limitations, errors, biases, and broader social impacts. Otherwise, we risk producing a generation that can use technology effectively, but cannot question it.

Sharifa Sultana is an Assistant Professor of Computer Science at the University of Illinois Urbana-Champaign and the Director of the Illinois Responsible AI Group.

Syed Ishtiaque Ahmed is an Associate Professor of Computer Science at the University of Toronto and an Associate Director at the Knowledge Media Design Institute.