US locks computers, China deploys AI cameras as countries battle exam cheating

Smart glasses are on display at a large supermarket in Seoul, Tuesday. Yonhap
Countries around the world are tightening exam security as cheating with generative artificial intelligence (AI) becomes increasingly sophisticated, introducing measures ranging from computer lockdown software and AI surveillance cameras to smartphone detectors and bans on smart glasses.
But even as authorities strengthen their defenses, students and developers continue finding ways around them, fueling an escalating technological arms race between exam administrators and would-be cheaters.
In the United States, the College Board, which administers the SAT college entrance exam, uses a dedicated testing application called Bluebook that locks down students' computers once an exam begins, preventing them from opening other programs or windows.
Since March, the organization has also explicitly prohibited smart glasses, even those worn for vision correction. Some overseas testing centers have introduced measures, with students' consent, that allow facial recognition and the monitoring of keyboard inputs and mouse movements.
Yet the restrictions have proved vulnerable.
According to a Nov. 6 report by The New York Times, the College Board discovered a $499 program called AutoSAT in March that could bypass Bluebook's restrictions, send exam questions displayed on the screen to an AI chatbot and display the correct answers.
At least four students used the program to take the SAT in March. The College Board responded by modifying its software code and filing a lawsuit against the developer.
China, meanwhile, has turned to AI to detect suspicious behavior during exams.
In June during the country's national college entrance examination, known as the "gaokao," authorities subjected students wearing glasses to additional checks and deployed AI-powered surveillance systems in testing rooms.
China Central Television (CCTV) said the systems could automatically identify unusual behavior among test takers and generate video clips of suspicious incidents for further review by exam supervisors.
Other countries are strengthening security before students even enter examination rooms.
India's National Eligibility cum Entrance Test (NEET), used for medical school admissions, prohibits large metal accessories and decorative belt buckles and requires all candidates to undergo physical searches before entering.
Turkey's university entrance examination similarly bans watches, jewelry and personal belongings.
Taiwan disqualified a candidate caught using smart glasses this year and plans to introduce stricter penalties for next August, in which merely possessing the devices, even without using them, will result in a score of zero for the subject concerned.
Authorities are also increasingly using technology to block or detect wireless communications.
India deployed 51,311 signal jammers at testing centers for a NEET examination held again in June following a question-paper leak.
In France, the education ministry has equipped all regional education authorities with mobile phone detectors during the written portion of the baccalauréat, the country's secondary school leaving examination, and randomly deploys the devices at examination sites.
Spain has adopted similar measures. During university entrance examinations held June 9 to 11, Catalonia conducted random inspections of testing rooms using detectors capable of identifying wireless signals from miniature earpieces and smart glasses.
Six public universities in Madrid also used the equipment for the first time during examinations in early June.
China's Ministry of Industry and Information Technology said it deployed mobile surveillance vehicles a cumulative 7,306 times and monitoring equipment a cumulative 15,624 times nationwide during this year's gaokao.
Despite these measures, cheating involving electronic devices remains widespread.
According to Ofqual, Britain's examinations regulator, 2,225 cases of cheating involving mobile phones or smart devices were recorded during secondary school examinations last summer, accounting for 44.3 percent of all student malpractice cases.
In France, 833 cases were referred to baccalauréat disciplinary committees last year. Of those, 55.04 percent involved smartphones, smartwatches or similar devices, while 8.5 percent involved AI-assisted cheating.
In Turkey, a student was arrested after attempting to use AI to answer questions during a university entrance examination in June. The student had reportedly concealed a mobile phone in his shorts, passed a security search and was caught after the examination.
The technological contest is also continuing in the United States. After the College Board modified Bluebook's code to block AutoSAT, the program's developers quickly released an updated version.
The persistence of such incidents has prompted growing calls for schools and examination authorities to rethink how students are assessed rather than relying solely on tighter security.
In a June article for The Conversation, Meena Jha, an associate professor at Australia's Central Queensland University, said memorization-based examinations were more vulnerable to cheating than assessments requiring critical thinking and problem-solving.
She said examination design itself needed to change to address the growing challenge posed by AI.
An internal committee at the Massachusetts Institute of Technology reached a similar conclusion in an August report, warning that AI detection tools could lead to an endless arms race with students.
Instead, the committee recommended combining multiple forms of assessment, including supervised in-person examinations and oral tests, to evaluate students' knowledge and abilities more reliably.
This article from the Hankook Ilbo, the sister publication of The Korea Times, was translated by a generative AI system and edited by The Korea Times.