AI driving steep rise in HSC cheating in schools

A steep rise in HSC cheating is being fuelled by artificial intelligence. More than 1000 incidents of misconduct were recorded in timed exams, take-home tasks and practical assessments last year.

Fresh data released by the NSW Education Standards Authority reveals for the first time how artificial intelligence is behind the spike of 1270 total cases of cheating in HSC assessment tasks recorded in 2025.

It is the second year the state has recorded more than 1000 incidents of cheating. Education analysts say the technology is “tempting” students who wish to outsource thinking.

Take-home assignments were the biggest area for misuse last year. Artificial intelligence was behind the 61 per cent of the 725 incidents of misconduct recorded.

There were 116 incidents where students tried to use an “electronic device” of some kind in a timed exam. That included sneaking in phones, smartwatches and wireless earbuds.

Principals earlier this year revealed students were going to extraordinary lengths to access artificial intelligence in exams. One student was caught cheating after carving a calculator case so it could fit their phone, including shaping a hole for the camera so they could take photos of the exam paper, the principal told The Sydney Morning Herald schools summit this year.

Macquarie University professor Matt Bower, who specialises in the innovative use of technology for learning purposes, said students graduating from school need an understanding of how AI can be used to improve learning and how to avoid cognitive offloading, which is using tech to reduce mental effort.

“Plagiarism and ‘ghost-writing’ are age-old problems, but generative AI has made it far easier and more tempting for students to outsource their thinking,” he said.

He said school systems must find the right balance between students developing the knowledge and skills pertaining to what they were studying and opportunities to develop AI literacy.

“We could also consider assessment tasks that require students to develop and demonstrate their AI literacies, including ethical and critical use of AI-generated content, as well as how to use AI to level up their creative outputs.”

The data showed the number of incidents of misconduct in official end of year exams, which comprise half their final mark, has remained small but has more than doubled in the past five years from 32 to 83 incidents last year.

NSW Secondary Principals’ Council president Ann Caro, whose organisation represents public school heads, said teachers had become more adept at identifying and detecting AI use, but she said many schools had changed their year 12 assessment tasks.

“It is important that the assessment reflects students’ capacity, not an artificial intelligence’s capacities,” she said.

That included adding in a viva voce, more teachers monitoring the progress of work before it was submitted for marking.

A major report earlier this year articulated how teachers fear students have become almost addicted to artificial intelligence. Some reported students resort to it to complete the most basic of tasks in class.

Numerous universities have come to the conclusion that it is futile to attempt to detect AI use in take-home tasks, moving to a “two-lane” model where certain assignments permit its use while others, typically in-person exams, strictly prohibit its use.

Diocese of Parramatta schools head of secondary learning Rebecca Birch said that unlike universities, schools were better placed to replace take-home exams with supervised in-class assessment tasks.

“Universities can’t do that at scale,” she said.

She said certain schools were moving towards a more constrained use of technology “and not the default to open the laptop” in a classroom.

“Some schools are moving towards low tech. There is more of a shift towards writing in books, which we know is also good for learning.”

According to the assessment misconduct data, English standard had the highest number of cheats last year: 191 incidents were recorded, followed by English studies and personal development, health and physical education.

A total of 532 students were caught using AI and 52 per cent received reduced marks; 39 per cent were awarded zero for the task; and 4 per cent received a warning.