AI Glasses Solved Four CSAT Math Questions in Ten Seconds
AI glasses solved four CSAT-style math questions in ten seconds in a Korea Times demo, days before Thursday's exam. Cheating tech is outpacing detection and reigniting reform calls.

Why it matters
- AI glasses solved four college entrance exam-style math questions in ten seconds in a Korea Times demonstration.
- Seoul education superintendent Jung Geun-sik has proposed abolishing the CSAT for 2040 admissions.
- This year's elementary-school intake fell below 300,000 for the first time, while four-year universities offer over 350,000 slots.
- Google's Gemma 3n can process images and text on-device without an external server, defeating internet blocks.
- Smart contact lenses under development in Dubai and Japan target some consumer applications for 2030.
AI glasses solved four college entrance exam-style math questions in ten seconds. That was the result of a recent demonstration by The Korea Times, and it lands days before Korea holds its annual College Scholastic Ability Test, the CSAT, on Thursday.
The demonstration matters because the CSAT is Korea's make-or-break standardized test. A single score can shape a student's university placement and, by extension, their career trajectory. If AI assistance can enter the exam room undetected, the fairness of that entire system comes into question.
But the ten-second demo also raises a larger issue than cheating. Even a perfectly secured test would leave the exam's educational purpose open to debate: what should students be learning in an AI-driven world, and what should a national exam actually measure?
How did the cheating actually happen?
The most urgent warning comes from a recent real-world case, not a laboratory. In the tax-accountant examination cheating case, a test-taker used AI together with a miniature camera and a smartphone to obtain answers throughout an exam without being detected.
The scheme never surfaced through proctoring. It surfaced because the cheater's unusually high score drew scrutiny. Had that score not raised suspicion, The Korea Times reports, the AI cheating may have gone unnoticed entirely.
That detail should concern exam authorities far more than any single gadget. Detection failed for the duration of the test. The fraud was caught by an anomaly in results, after the fact.
Notably, the tax-accountant setup did not even involve AI glasses. It used equipment already available to ordinary consumers. The devices now drawing education authorities' attention ahead of Thursday's CSAT — smart eyewear — represent the next step, not the first one.
Why can't authorities just block the technology?
Because the technology keeps getting harder to see, hear, and block. Three developments illustrate the trajectory:
- Smart contact lenses are under development in Dubai and Japan, with some consumer applications targeted for 2030. Authorities may eventually have to look beyond the glasses students wear to the lenses on their eyes.
- On-device AI, such as Google's Gemma 3n, can process images and text without connecting to an external server. This means even strict internet restrictions inside exam halls cannot prevent its use.
- Silent-speech interfaces already exist in research form. Cornell's EchoSpeech recognizes a limited set of silently mouthed commands using sensors on glasses, while InvisiSpeech detects subtle mouth vibrations with a sensor mounted around a back molar.
None of these are ready-made cheating systems today. The Korea Times is careful on this point, and the distinction matters. But together they point to a future in which AI assistance becomes progressively invisible to proctors — no screen, no phone in hand, no network traffic to intercept.
Each countermeasure closes one door while the technology opens another. Security fixes may buy time, but they do not settle the underlying argument about what the exam is for.
What would a meaningful exam actually test?
Here the story shifts from surveillance to pedagogy. Independent problem-solving still matters, the reporter notes, and for a specific reason: students need knowledge of their own to judge whether an AI-generated answer is accurate or evidence-based.
That standard cuts both ways. If students must retain enough expertise to evaluate machine output, then the exam itself should reward more than picking the right option under time pressure. The Korea Times argues for giving greater weight to:
- explaining a solution,
- spotting a less obvious error, and
- applying knowledge to an unfamiliar problem.
A multiple-choice format optimized for speed and standardization sits awkwardly beside those goals. The CSAT's structure was built for an era when the main threat to fairness was a hidden cheat sheet, not a networked — or fully offline — intelligence whispering answers.
Who is calling for CSAT reform?
The reform camp now includes a senior official. Seoul education superintendent Jung Geun-sik has proposed moving toward absolute grading and written-response assessments, with the eventual abolition of the CSAT for 2040 admissions.
That timeline is not arbitrary. Demographics are forcing the issue from the other direction. This year's elementary-school intake fell below 300,000 for the first time. Four-year universities alone offer more than 350,000 admission slots.
The arithmetic is stark. Seats already outnumber children entering school. By the time this year's elementary intake reaches college age, demographic change will have forced a reckoning on how — and whether — a hyper-competitive national exam still makes sense.
AI cheating and a shrinking student population thus push in the same direction. Both strengthen the case for accelerating reform that might otherwise stall for another decade.
Why hasn't Korea reformed the exam already?
Because trust is the hardest credential to earn. As past attempts have shown, reform will not be easy. Any alternative assessment must earn the trust of students and parents who are rightly concerned about admissions fairness.
The CSAT, for all its rigidity, has one great asset: it is perceived as objective. A standardized, anonymously scored test is hard to game through connections, wealth, or subjective judgment. Replace it with written responses and absolute grading, and families will ask who scores the essays and by what criteria.
Consensus on what should replace the current system remains elusive, The Korea Times reports. That gap between dissatisfaction and agreement has delayed reform for years.
The irony is sharp. The same technology destabilizing the exam may be the thing that finally forces the conversation. By calling the future of standardized testing into question, AI may force a debate that has been delayed for too long.
What happens next?
In the short term, expect a security escalation around Thursday's CSAT, with attention focused on smart glasses and similar wearables. That is the necessary, immediate response to keep the exam fair and credible for this year's test-takers.
The longer term is less certain. Smart contact lenses targeted for 2030, on-device models that need no network, and silent-speech sensors all suggest the surveillance approach has a shelf life. Jung's 2040 abolition proposal and the sub-300,000 elementary intake give Korean education policy two deadlines moving toward each other.
By the time they meet, Korea will have had to answer the question AI raised in that ten-second math demo: not how to keep machines out of the exam room, but what an exam is supposed to prove when the machines can pass it.
Source: feed.koreatimes.co.kr
More from Elena Vasquez
Show full bio
Market editor covering technology and product news in artificial intelligence and machine learning at AI In Context.
708 articles
Related articles
- Korea adds 8,400 staff for AI-cheating safeguards at CSAT
- South Korea Cracks Down on AI Cheating Before Thursday’s CSAT
- Korea to Cut Wi-Fi at College Entrance Exam Sites Over AI Cheating
- Smart Glasses Banned as South Korea Shields Suneung From AI Cheating
- Top Tax Exam Scorer Admits Using Miniature Camera and AI to Cheat