HEC AI rules for students face a major challenge over tool disclosure

The issue has raised questions about how student work should be assessed.

HEC introduces new AI measures

The Higher Education Commission has introduced two major AI measures for universities. It has made a three-credit-hour AI course compulsory for undergraduate and postgraduate students from Fall 2026 and has also issued a draft policy on generative AI use.

The policy says students should not submit AI-generated work as their own. It also asks students to disclose which AI tools they used and explain why they used them.

The main concern is whether universities can actually check these declarations. The draft itself says students should not be punished only because an AI detection system reports that their work was AI-generated.

Why AI skills still need teaching

The debate is not only about students using AI to cheat. A bigger question is whether universities are teaching skills that AI can now perform quickly, such as data cleaning, forecasting and some coding tasks.

However, AI does not remove the need for subject knowledge. Students still need to understand which question to ask, whether data is reliable and whether an apparently good answer is actually correct.

The argument is similar to teaching arithmetic even though calculators are available on every phone. Education is meant to build skills and thinking ability that students cannot simply hand over to a machine.

The compulsory AI course could become weak if it only teaches students how to write prompts for chatbots. Students already have free access to such tools, so universities may need to focus on deeper and more useful AI skills.

What universities could teach instead

Students could learn how AI systems work in practical settings, including APIs, structured data and automated tasks. This would help them use AI to handle large amounts of information instead of simply chatting with a model one document at a time.

The difference is important because using an AI tool and building something with it are not the same skill. A graduate who understands how to connect AI to a larger system can potentially automate tasks across thousands of records.

The source also points to software developer Peter Steinberger as an example of how human knowledge and AI can work together. His technical experience and ideas were important, while AI tools helped him build faster.

This suggests that AI can multiply existing ability rather than create expertise from nothing. For universities, the challenge is therefore to give students enough knowledge and practice before allowing AI to take over difficult parts of their learning.

A different way to check student work

One possible solution is to separate learning from grading. Students could use AI openly for take-home practice, while important grades could come from supervised work, presentations, vivas and short oral discussions.

A student who understands an assignment should normally be able to explain it when questioned. This approach could provide a more direct way to judge learning without relying completely on AI detection software.

Supervised written exams could remain the main method, especially for large classes. Universities could focus less on questions based only on memory and more on questions that require students to understand and apply what they have learned.

Oral checks could be used for projects, theses and group assignments. Students could be asked to explain parts of their submitted work so teachers can see whether they actually understand it.

Why AI detectors remain a concern

AI detection software can make mistakes because it judges the final document rather than seeing how the work was produced. The source cites a 2023 Stanford study in which several detectors wrongly identified many essays by non-native English speakers as AI-generated.

The source also notes that some companies have since changed their systems and reported improved results. However, it says there has been no validation of such systems specifically on Pakistani student writing.

Because of this uncertainty, detector scores should not be treated as proof that a student used AI. Universities should first test such systems on local student writing before relying on them for academic decisions.

The bigger issue is not whether students use AI. It is whether universities can still measure what students actually learn.

AI is changing the value of written assignments, but it has not removed the need for education. Students still need to develop judgment, knowledge and the ability to explain and defend their work.

The compulsory AI course could be useful if it teaches students how to build and work with AI systems. If it becomes only a course about writing prompts, it may quickly lose its value.

The most practical approach may be to allow AI where it helps learning but assess students in settings where their own knowledge can be seen. In simple terms, universities need to put education first and AI second.

Also read: HEC offers Sri Lanka IT scholarship for Pakistani students

Also read: HEC opens overseas scholarships for students targeting top global universities