Smartphone bans as a blueprint for school AI policy
Age-tiered access and transparent procurement as the foundation for AI use standards
Evidence-based evaluation and teacher-led governance to strengthen learning focus and fairness

This article is a reconstruction tailored to the Korean market based on a contribution to the SIAI Business Review series published by the Swiss Artificial Intelligence Institute (SIAI). The series aims to present researchers’ perspectives on the latest issues in technology, economics, and policy in a manner accessible to general readers. The views expressed herein are those of the author and do not necessarily reflect the official position of SIAI or its affiliated institutions.

Forty percent of education systems worldwide already restrict smartphone use in schools. As of late 2024, 79 countries had implemented laws and policies that ban or limit smartphone use on campus. This shift illustrates how societies absorb and regulate new technologies. The pattern is consistent: adoption comes first, impacts and side effects are assessed, and rules are then established to protect students. The experience forged around smartphones now needs to be applied to large language models (LLMs).

Both technologies are deeply embedded in everyday life, and both are designed to capture and hold attention. The result has been diminished student concentration and weaker learning immersion. The question now turns to how quickly and clearly AI policy can be built, drawing on the smartphone playbook.

Lessons Left by Smartphones

Restrictions on smartphone use are increasingly viewed as a tangible outcome of education policy responding to technological change. UNESCO has documented how scattered guidance across countries evolved, over a short period, into national-level bans or restrictions. The number of countries implementing such policies expanded from 60 in 2023 to 79 by late 2024.

The policy shift moved fast. The United Kingdom, in February 2024, issued national guidance enabling headteachers to restrict mobile phone use during class. The Netherlands banned use in classrooms except for learning or accessibility purposes and recorded high compliance soon after implementation. Finland, starting in 2025, will restrict device use during lessons and grant teachers the authority to confiscate devices that disrupt learning.

These measures reshaped the learning environment. The prevailing rule became keeping devices unrelated to learning or health outside the classroom, which translated into higher student focus and the restoration of teachers’ instructional authority. According to PISA 2022, roughly two-thirds of OECD students reported feeling distracted by smartphones during math class, and their average scores were lower than those of students who did not report such distraction. In Australia, one year after nationwide restrictions, 87% of 1,000 school administrators in New South Wales said in-class distraction had declined, and 81% said learning outcomes had improved.

Estonia, however, chose “controlled device use” and AI literacy education instead of a ban. Rather than limiting devices outright, it adopted an approach that places teachers at the center and integrates technology into instruction. UNESCO assessed such efforts positively while also noting that research and evidence remain insufficient relative to the speed of technological change. The core issue ultimately lies in moving beyond a ban-permit binary and establishing clear boundaries that enable safe, education-aligned use of technology.

Direction for Designing School AI Policy

Bringing AI into school education will stand as a central task for future education policy. The experience with smartphone policy shows that effective regulation needs to remain simple while staying grounded in operational reality. Exceptions can be recognized for educational necessity such as learning objectives, health, and accessibility, while data protection needs to be strengthened.

The risks introduced by AI extend beyond attention erosion. Student data may be collected indiscriminately, and excessive reliance on AI may take root. The risk of misinformation affecting evaluation or grades is also substantial. In 2024, 70% of U.S. teenagers used generative AI, and 40% used it for learning. In the same year, a Pew Research Center survey found that 25% of U.S. K–12 teachers said AI harms education, while 33% assessed positive and negative effects as roughly balanced. These findings show a widening gap between rapid AI diffusion and the still-thin standards for response and use in schools.

Policy design should be refined around age-tiered access rules. Students under 13 should be limited to dedicated tools managed by schools, with chat content set to avoid being stored. Ages 13–15 should be allowed access only via school accounts, with usage logs and content filtering made mandatory. Students 16 and older should be permitted to use AI only under teacher supervision, with clear disclosure of purpose and sources.

Governance standards are also required at the adoption stage. During contracting, schools should configure settings so that minors’ account data does not automatically persist, and age verification should be conducted in a way that does not collect personal information. Teachers or administrators should be able to disable AI chat functions or extensions when needed, and vendors should disclose reports detailing data-handling practices and safety-validation results.

Evidence and Evaluation for AI Policy

Unlike smartphones, evidence on AI’s learning impact remains limited. With smartphones, the association between distraction and lower academic achievement is clear, while definitive proof that bans directly raise grades remains absent. Even in Australia, positive change was reported, yet much of it rested on surveys of school administrators.

AI carries greater uncertainty. As of fall 2024, 48% of U.S. school districts had implemented AI teacher training, doubling within a year, and 72% of teenagers were using AI conversational services. More than half of students said they use them regularly. This points to risks that extend beyond learning tools toward emotional dependency or distorted relationships. Policy needs to draw a clear line at this point.

The research base is still in an early phase. PISA 2022 surveyed 690,000 15-year-old students across 81 countries, and the 59–65% distraction share reflects the OECD average. Common Sense and RAND materials are sample-based U.S. surveys that remain limited to trend detection. The cases of the United Kingdom, the Netherlands, and Finland also remain in early implementation stages.

Accordingly, AI policy needs to place its center of gravity on evaluation. Every school district should include a clear evaluation plan when adopting AI and should continuously track outcomes such as teacher workload reduction, plagiarism reduction, and learning gains for low-income and disabled students. Results should be disclosed on a regular basis, and trust in the field needs to begin with transparent evaluation.

The Path Schools Need to Take

The experience of smartphone regulation showed what is required when technology enters schools. Schools have already learned that introducing technology without clear rules can destabilize instructional quality. That lesson applies to AI as well.

Generative AI is spreading rapidly across all subjects and learning processes. Ambiguous standards can produce divergent interpretations across schools and widen access gaps among students. Simple and consistent policy can raise learning focus, protect fairness in assessment, and help students use technology appropriately. What is required now is execution. Age-tiered access standards, transparent management of usage logs, clear procurement procedures, and systematic evaluation need to form the core of policy. Once these principles take root, schools can sustain learning quality and fairness even amid accelerating technological diffusion.

Please refer to the original research article, From Smartphone Bans to AI Policy in Schools: A Playbook for Safer, Smarter Classrooms. The copyright for this article belongs to the Swiss Artificial Intelligence Institute (SIAI).

A shift toward reasoning- and evidence-based writing education in response to AI proliferation
The need to establish transparent assessment frameworks that incorporate the process of AI use
Strengthening teacher capacity and ensuring a fair learning environment as core challenges

This article is a reconstruction tailored to the Korean market based on a contribution to the SIAI Business Review series published by the Swiss Artificial Intelligence Institute (SIAI). The series aims to present researchers’ perspectives on the latest issues in technology, economics, and policy in a manner accessible to general readers. The views expressed herein are those of the author and do not necessarily reflect the official position of SIAI or its affiliated institutions.

The share of U.S. teenagers using ChatGPT for academic purposes nearly doubled within a year, rising from 13% to 26%. This is not a fleeting trend but a signal of a fundamental shift in learning methods. Spreading across grade levels and backgrounds, this change indicates that classroom writing is no longer an individual endeavor, but a process increasingly shaped through collaboration between humans and artificial intelligence.

If AI writing education ignores this reality, students will repeatedly face an irrational situation in which they are penalized for using tools that are becoming essential for future learning and work, while those who conceal their use are rewarded. The issue is not whether AI should be permitted, but what competencies students must develop through its use.

Redefining Writing Education

In the past, writing was perceived as an isolated task that individuals had to complete alone. Today, however, writing has evolved into a collaborative process in which humans and artificial intelligence jointly develop ideas and refine expression. The focus of writing is shifting away from sentence-level polishing toward the ability to reveal the flow of thought and substantiate arguments.

AI writing education must redefine the meaning of literacy. At its core are the abilities to develop questions into evidence-based claims, analyze and verify sources, collaborate with AI to construct drafts, and clearly articulate one’s own reasoning. To achieve this, more instructional time should be devoted to teaching argument design rather than grammar correction. Structuring claims and evidence, systematic note-taking, and brief oral presentations are concrete methods. Students must learn the principle that while AI can be an efficient tool, judgment and responsibility remain human obligations.

These changes are already being detected in educational settings worldwide. Recent policy directions have shifted away from banning AI toward establishing principles for human-centered use. Yet many countries still lack clear guidelines, and schools have been slow to adopt validated educational tools. The direction, however, is clear: use AI, but manage the thinking process transparently.

A Shift Toward Assessment-Centered Reform

If assessments favor students who conceal AI use, students will do exactly that. If, instead, assessments emphasize transparency of reasoning and evidence, students will disclose their process and clearly present their sources. Accordingly, AI writing education must redesign assignments around verifiable evidence and clarity of expression.

Assignments should move away from formulaic essays toward engagement with real-world issues. Examples include analyzing local budget expenditures, comparing and evaluating policy reports, or re-examining existing arguments using newly released statistical data. Tasks that require students to present their proposals in short presentations of around three minutes, based on their own research, are also effective.

All assignments should require the submission of a process report. This report must include the prompts used, the scope of AI involvement, the sources consulted and their verification procedures, and a methodology note of approximately 100 words. While the completed text evaluates outcomes, the report assesses the student’s thinking process. Evaluating both together ensures transparency of learning and academic integrity.

The methodology note should be concise but specific. For example: “I used an AI assistant to draft an outline, and completed paragraphs two through four after receiving advice on structure and tone. Statistical data were cross-checked against original datasets from the Pew Research Center and the Organisation for Economic Co-operation and Development, with errors corrected after two rounds of verification.” The goal is to enable teachers to move beyond surveillance and instead assess the quality of reasoning and evidence.

This approach also aligns with student demands. In a 2024 survey of 3,839 university students across 16 countries, 86% reported already using AI for academic work, yet 58% felt they lacked sufficient understanding of AI, and 48% believed they were unprepared for an AI-centered work environment. This gap is precisely what new writing education must address. A system that teaches and evaluates verification, disclosure, and context-appropriate expression is required.

The Limits of Detection Technologies

More schools are relying on detection tools to curb AI misuse, but this approach fails to address the fundamental issue. Even developers acknowledge accuracy limitations. OpenAI discontinued its text classifier due to low reliability, and major universities and educational institutions advise against using detection results as grounds for disciplinary action. Turnitin, a widely used plagiarism detection software, also withholds AI detection results when confidence levels fall below 20% in order to reduce false positives.

The limitations of detection tools are evident. There are cases in which carefully polished student writing is mistakenly identified as AI-generated. Detection may serve as a reference, but it must never stand alone as a basis for judgment. Education policy must reflect this reality.

Moreover, directly using detection results for disciplinary measures poses legal and ethical risks. Cases in which students were sanctioned based on ambiguous rules and low-reliability technologies have already escalated into legal disputes. This erodes trust among teachers, students, and administrators and generates unnecessary conflict. Sanctions must therefore never rely solely on detection results, and evidence must always include documentation of the writing process and source verification.

Strengthening Teacher Capacity and Educational Equity

Teachers must acquire new instructional skills, including prompt design, source verification, and process-based assessment. Despite ongoing teacher shortages, such retraining is no longer optional. Educational authorities’ medium- and long-term plans increasingly identify not only teacher recruitment but also the capacity to perform these new roles as a core task.

Teachers are now expected to use AI as a supplementary educational tool. Their role includes co-designing writing processes and providing feedback centered on logic and evidence. Examples include demonstrating verification procedures in real time during class or conducting brief oral presentations in which students explain why they chose specific claims and sources. These instructional methods prioritize reasoning over mere replication and reduce AI misuse.

Ensuring a fair learning environment is equally critical. Paid AI services must not become instruments that widen educational gaps. Schools should institutionally support access to reliable, shared AI tools. Clear disclosure standards are also needed to ensure that non-native speakers are not penalized simply for using grammar correction functions.

The core of AI utilization education lies in teaching appropriate use. AI may be employed to expand ideas or verify data, while organizing thoughts and refining expression should remain student-driven. This distinction is not about replacing writing training, but about concentrating on essential competencies.

The Direction of Education in an Era of Change

The purpose of AI writing education is not control, but capacity building. Proper use of artificial intelligence can clarify thinking processes, enhance expressive quality, and improve fairness in assessment. Students must learn how to formulate claims, verify evidence, and rationally leverage available assistance. Even without reliance on detection technologies, academic integrity can be ensured by jointly evaluating process and outcome. Schools must establish accredited tools and clear guidelines to narrow gaps between teachers and students, while strengthening data protection frameworks.

If education remains anchored to traditional writing standards that exclude AI, it will fail to keep pace with reality. Future writing education must place evidence and verification at its core and advance through transparent processes that build trust.

For reference, the original version of this research article is AI Writing Education: Stop Policing, Start Teaching. The copyright of this article belongs to the Swiss Artificial Intelligence Institute (SIAI).

U.S. H-1B Employment Visas Face a “$100,000 Threshold”
Erosion of U.S. Corporate Global Competitiveness Pushes High-End Jobs Offshore
Intensifying Global Race to Attract Talent as Visa Tightening Proves Self-Defeating

Dire forecasts are proliferating over President Donald Trump’s sweeping hike in H-1B visa fees. Projections suggest U.S. work permits could decline by more than 5,000 each month, severing a vital growth platform for startups. Concerns are mounting that the policy could accelerate offshoring, driving advanced jobs overseas. As American firms reel from diminished access to the global talent pool, countries such as the UK and China are moving aggressively to seize the opportunity and lure skilled workers.

Work Permits to Fall by as Many as 5,500 a Month, Capital Strength Divides Firms

According to Bloomberg on the 29th (local time), JPMorgan warned in a recent report that the White House’s imposition of a $100,000 H-1B visa fee could slash immigrant work permits by up to 5,500 per month. “While the overall impact on the U.S. labor market may be limited, the shock will be profound for IT companies and Indian immigrants,” the report noted.

In fiscal year 2024, roughly two-thirds of approved H-1B applicants were in computing roles, with half concentrated in professional, scientific, and technical services. Indian nationals accounted for 71% of all approvals. Bloomberg reported that of the 141,000 new H-1B petitions filed last year, around 65,000 were processed abroad — a category JPMorgan expects to be directly hit by the fee. Analysts added that U.S. firms reliant on foreign skilled labor could lose as many as 140,000 jobs annually, over 10,000 a month.

The exorbitant visa fee is expected to cleave the IT sector into firms with strong financial reserves and those without. The Wall Street Journal observed that the $100,000 charge would function as a “selective barrier” in Silicon Valley. Some Big Tech leaders expressed guarded optimism. Nvidia CEO Jensen Huang and OpenAI’s Sam Altman struck a positive tone in a joint CNBC interview, while Netflix chairman Reed Hastings posted on X calling the measure a “great solution,” arguing it would deter lower-value visa applications. Parker Conrad, CEO of software firm Rippling, said he would willingly pay the fee if it guaranteed access to visas.

Startup founders, however, voiced stark opposition. Critics argue the Trump administration is erecting a wall that only well-capitalized corporations can scale. Greg Morrisett, dean of Cornell Tech, warned, “Large firms can absorb the added cost, but for startups it is devastating. This fee hike will stifle new ventures.” He underscored the program’s importance by noting that “Microsoft, Google, and Meta all grew through H-1Bs.”

UK and Germany Step Up Foreign Talent Recruitment

Concerns are rising that the visa fee regime will accelerate offshoring, with key nations poised to capitalize. Europe is at the forefront. The Financial Times reported that UK Prime Minister Keir Starmer is weighing a plan to scrap visa fees for top global talent.

Barney Hussey-Yeo, founder of British fintech unicorn Cleo, openly courted U.S. talent on LinkedIn, offering to assist those forced out by H-1B restrictions. Cleo reportedly posted around 100 such job ads through its London office, explicitly targeting professionals affected by the U.S. policy. “We know being forced out of the U.S. was not your plan,” one posting read, “but sometimes the best opportunities come from unexpected change.”

The Startup Coalition, a British policy group, urged the government to “seize the opportunity” created by Trump’s clampdown and implement policies to redirect talent flows toward the UK.

Germany is also framing the U.S. visa shock as a “golden opportunity.” Bernhard Rohleder, head of the German digital industry association Bitkom, said the American policy “could help Germany and Europe attract top-tier talent.” Philipp Ackermann, Germany’s ambassador to India, promoted his country’s immigration policy as “reliable, modern, and predictable,” stressing job security for Indian workers. He noted that Indians working in Germany on average earn higher incomes than German nationals, bolstering their contribution to society and welfare systems.

Canada Seeks Redemption, China Escalates Talent Offensive

Canada — once derided as “America’s 51st state” — is now mobilizing to capture talent fleeing the U.S. Ottawa has integrated the H-1B turmoil into its immigration strategy, aiming to absorb global tech workers. Cities such as Vancouver, Toronto, and Montreal have already emerged as R&D hubs for major U.S. tech firms. Vancouver, in particular, has served as a fallback labor pool for Seattle-based giants like Amazon and Microsoft. Analysts predict America’s harsher immigration stance could transform into Canada’s brain gain.

China, meanwhile, is rolling out an even more aggressive policy. According to the State Council, Beijing will launch a new “K Visa” for STEM talent starting next month. Unlike conventional visas, it does not require prior job offers or research appointments, allowing foreign applicants to enter China for study or job hunting — and remain even without employment.

Foreign Ministry spokesperson Guo Zhaokun said the new category was created “to promote youth science and technology talent exchanges between China and abroad,” with details to be released by embassies and consulates. State-run Global Times added that K Visa holders could engage in education, cultural, and scientific exchanges, as well as corporate and business activity.

Although the plan was unveiled in August, the timing of its rollout — coinciding with Washington’s fee surge — has magnified its significance. Analysts argue the Trump administration’s move is a “self-defeating policy that destabilizes critical talent pipelines,” giving China a chance to present the K Visa as an alternative. Reuters called the move “perfectly timed,” noting that as H-1B applicants search for substitutes, China’s new visa could tilt the global talent race in Beijing’s favor.

Classroom Temperature Has a ‘Material Impact’ on Academic Performance
Digital Device–Driven Distraction Remains a ‘Persistent Problem’
Cooling Cannot Serve as a ‘Panacea’

This article is a reconstruction tailored to the Korean market based on a contribution to the SIAI Business Review series published by the Swiss Artificial Intelligence Institute (SIAI). The series aims to present researchers’ perspectives on the latest issues in technology, economics, and policy in a manner accessible to general readers. The views expressed herein are those of the author and do not necessarily reflect the official position of SIAI or its affiliated institutions.

As global warming intensifies, schools are rushing to install air-conditioning systems. A Japanese study shows that on days when temperatures exceed 34 degrees Celsius, students taught in non-air-conditioned classrooms see their test scores fall by 0.56 standard deviations. When air conditioning is introduced, however, 73% of the lost performance is recovered.

Classroom Cooling and Its ‘Substantial’ Impact on Academic Outcomes

Last year, Japan equipped most public schools with air-conditioning. This marks a striking shift from 2006, when the installation rate stood at just 10%. Despite this progress, student academic performance in Japan and other countries continues to decline, indicating that factors beyond cooling are at play. Between 2018 and 2022, the average mathematics score across OECD countries fell by roughly 15 points. In South Korea, one of the top-performing nations, scores dropped sharply from 554 in 2012 to 527 in 2022.

Policymakers have responded swiftly. New York State in the United States has codified an indoor temperature cap of 31.1 degrees Celsius, requiring schools to implement emergency measures when the threshold is exceeded. The United Kingdom, while avoiding a legally binding limit, relies on guidelines focused on risk assessment and comfort. Such measures clearly improve health outcomes and reduce discomfort, but they do not appear to automatically translate into higher test scores.

Instruction, Curriculum, and Attention Remain ‘Critical’

Heat does have a direct effect on learning. Research indicates that rising temperatures disproportionately harm the academic performance of low-income and minority students. Those already struggling to keep up with coursework face even greater challenges. In these circumstances, air conditioning, while imperfect, alleviates a significant share of the problem. A Japanese study that tracked individual student outcomes and compared them with local weather patterns provides evidence of this effect.

At the same time, the research also shows that investments in cooling infrastructure do not mechanically boost academic achievement. The quality of instruction, rigor of the curriculum, attendance rates, and time spent focusing on assignments remain decisive factors. OECD analyses similarly conclude that while the learning environment matters, it cannot substitute for effective teaching methods and structured study practices. In short, policies centered solely on cooling enhance comfort and safety but fail to raise academic performance on their own.

Cool, Clean Air and Its Contribution to ‘Focus’

Cooling does help reduce cognitive load—the mental effort required to process information and complete tasks—while lowering fatigue and improving attendance. High-quality HVAC systems have been shown to reduce absenteeism and suspensions, and to contribute marginally to gains in reading and mathematics scores. Cooling and ventilation mitigate thermal stress and support concentration. Yet even after accounting for these benefits, the greatest enemy of learning remains distraction.

According to the 2022 Programme for International Student Assessment (PISA), students who reported being distracted by classmates’ mobile devices during mathematics lessons scored an average of 15 points lower. When digital device use is curtailed in a disciplined manner, scores rebound, suggesting that the core issue lies not in the devices themselves but in the distraction they generate. No matter how cool the classroom, maximizing learning outcomes ultimately requires teachers who foster concentration and structured routines.

A ‘Multifaceted Approach’ Linking Heat and Instruction

Even schools in advanced economies with state-of-the-art cooling systems struggle with breakdowns, noise, and power constraints. Standards vary widely, including in countries like the United States, which lacks a federal indoor temperature limit. Budget shortfalls, volatile electricity costs, and climate targets compound these challenges.

Addressing heat therefore calls for a multifaceted strategy. Aligning class schedules with temperature patterns—placing cognitively demanding subjects during cooler periods and lighter coursework during hotter hours—offers one practical option. New York State’s 31.1-degree benchmark, combined with such scheduling, could help preserve students’ capacity to focus on assignments.

More broadly, upgrading HVAC systems should be treated as an element of education policy. Emphasizing filter maintenance, air circulation, carbon dioxide monitoring, and regular servicing can improve comfort and academic outcomes without large-scale renovations. Restricting digital device use during key instructional periods would further enhance student focus.

Rigorous measurement is also indispensable. Schools that upgrade cooling and ventilation systems should track attendance, behavioral incidents, and academic performance, benchmarking results against comparable institutions. The ultimate objective is not cooling itself, but higher academic achievement.

Cooling can mitigate the adverse effects of heat on learning, but it is no cure-all. Student focus, curriculum design, and the overall management of the classroom environment remain enduring priorities.

For the original version of this research article, please refer to Cooler Classrooms, Cooler Heads? Why Air Conditioning Alone Won’t Fix Learning, Copyright of this article belongs to the Swiss Artificial Intelligence Institute (SIAI).

The spread of AI devices and connected toys into infant development environments
The core challenge lies in preventing data misuse while establishing design standards that promote interaction
Institutionalization across policy, education, and households, alongside equity safeguards, is essential

This article is a reconstruction tailored to the Korean market based on a contribution to the SIAI Business Review series published by the Swiss Artificial Intelligence Institute (SIAI). The series aims to present researchers’ perspectives on the latest issues in technology, economics, and policy in a manner accessible to general readers. The views expressed herein are those of the author and do not necessarily reflect the official position of SIAI or its affiliated institutions.

Artificial intelligence is rapidly permeating the environments in which infants grow and develop. In the United States, one-third of the population aged 12 and older owns a smart speaker, and many households install such devices throughout the home, exposing infants even as they sleep, babble, and play. The market for connected toys is also expanding rapidly and is projected to grow into a tens-of-billions-of-dollars industry within the next decade. AI has already embedded itself in the lived reality of infant development.

The critical issue is direction. Society must decide whether to permit devices that collect data while weakening human interaction, or to establish regulations and standards that strengthen communication between parents and children. During the first 1,000 days of life, the highest priority must be immediate conversation and physical interaction between parent and child.

Evaluation Standards for AI Toys

The key issue is not simply whether a device uses artificial intelligence. What matters is whether it replaces the role of parents or facilitates interaction. Language and cognitive development in early childhood depend on the rapid back-and-forth exchanges known as “serve-and-return” interactions. Research consistently shows that the frequency of conversational exchanges between parents and children has a greater impact on language development and later achievement than the sheer number of words a child hears. This has been demonstrated across multiple methodologies, including audio-recording observations and neuroimaging studies. Accordingly, policy attention should focus not on AI that talks directly to babies, but on tools that increase interaction between parents and children.

This principle also explains why video viewing continues to have negative effects on infant development. Pediatric guidelines recommend that children under 18 months avoid screen viewing, except for video calls, and that even older infants be limited to high-quality content viewed together with adults. Neurophysiological research supports the same conclusion. Infants learn best through voices that respond to their sounds, faces that move in synchrony with expressions, and hands that jointly manipulate objects. No matter how sophisticated, video cannot provide this kind of reciprocal responsiveness. Devices that cannot demonstrate the benefits of responsive human interaction should not be incorporated into infants’ daily lives.

Necessary Safeguards

The debate surrounding AI toys is not about blanket permission or prohibition. The core task is to establish standards that distinguish desirable designs from problematic ones. Only then can parents, educators, manufacturers, and policymakers operate within an environment of trust.

First and foremost, data protection is critical. Some smart toys have transmitted children’s behavioral data to corporate servers without encryption. For infant products, this constitutes a serious risk. The United States has strengthened regulations on children’s personal data, restricting commercial use without parental consent, while Europe has banned targeted advertising aimed at minors. Although not all issues have been resolved, minimum standards are now in place. Devices that collect data without authorization, create opaque user profiles, or are designed to maximize usage time should no longer be acceptable.

Verification of effectiveness is also essential. Claims of improved language development must be supported by independent research. Outcomes should be assessed using objective indicators such as the frequency of parent–child conversations, standardized developmental assessments, and measures of sleep or stress. Studies must be conducted at sufficient scale and with rigorous methodologies.

Toy design should prioritize shared play. AI toys for infants should be structured to respond to a child’s sounds in ways that prompt parental engagement, and they should automatically shut down when a caregiver is not present. Data processing should, in principle, occur on-device, with clear disclosure of collection purposes and retention periods. Evaluation criteria should emphasize not usage time, but how often and how effectively parents and children communicate.

Equity considerations must also be addressed. Economically advantaged households can more easily access educational toys, speech therapy, and parental education, while less affluent families face persistent disadvantages. If AI toys remain premium products, or if free offerings rely on data extraction as compensation, disparities will widen further. Public spaces such as libraries and health centers should provide access to validated products alongside parental education programs.

What We Should Build

Future AI toys should serve not as substitutes for parent–child interaction, but as complements that reinforce it. The essential goal is to encourage adults and children to converse more frequently and engage in shared play.

For example, a screen-free doll could be equipped with sensors and voice recognition. When an infant grasps a block and makes a sound, the device could signal a nearby caregiver. Once the caregiver responds, the device would cease operation, avoiding disruption of the conversational flow. Parents could later review interaction patterns and research-based guidance through an app.

Applications are also possible in early childhood education settings. During reading sessions, a small device could recognize specific words and suggest questions to teachers. Rather than reading the book itself, the device would help teachers engage children in deeper dialogue. Some studies have shown that such tools increase children’s questions and participation.

Policy and Practice

Educational settings must move beyond the mere introduction of devices to establish clear usage standards. Childcare programs should integrate language-centered interaction into teacher training, ensuring that AI tools function as supplements to, rather than drivers of, communication. Over time, the goal should be to strengthen teacher involvement rather than allow devices to dominate interactions.

Parents also require clear guidance. Products without proven effectiveness cannot be trusted, and devices that require separate registrations or excessive data permissions to function are difficult to regard as child-centered designs. For infants, cutting-edge features are irrelevant. What matters is fostering natural growth in language and emotional connection in everyday life. Beneficial AI should remain in the background, with the parent’s voice at the center.

Policy must institutionalize these principles. The United States has reinforced child privacy regulations to limit data commercialization without parental consent. Regulatory frameworks in Europe and the United Kingdom likewise emphasize the best interests of the child and data minimization. Enforcement authorities should prioritize connected toys and infant monitoring devices for scrutiny, extending beyond formal inspections to include substantive technical reviews and data analyses. When corporate claims and data are rigorously examined, products can be guided toward designs centered on shared play and safety.ㅁ

Future Challenges

Voice-based devices, like radios and smartphones before them, are now embedded in everyday life. The crucial question is how society chooses to use this technology and what roles it assigns. Privacy regulations must be strengthened to block excessive data collection, and any claims regarding developmental benefits must be backed by evidence. Devices should be developed that do not disrupt parent–child interaction but actively promote it, accompanied by policies that ensure equity by prioritizing support for families most in need.

The objective of policy choices is clear: to develop and disseminate technologies that enrich conversation, contact, and trust during infancy.

For the original version of this research article, please refer to Not Guinea Pigs, Not Glass Domes: How to Design AI Toys That Help Babies Learn. Copyright of this article belongs to the Swiss Artificial Intelligence Institute (SIAI).

China’s youth unemployment rate jumped to 18.9% in August
Seasonal factors compounded by an economic slowdown and flawed policy design have converged to intensify labor market stress
“With prospects like this, maybe the military is an option,” young Chinese increasingly scour the job market in search of stable employment

China’s youth unemployment rate surged last month as the economy slowed and hiring at private firms dropped sharply. The downturn, compounded by seasonal factors and other unfavorable conditions, has led to a marked deterioration in labor market indicators.

China’s Youth Employment Crisis Deepens

On the 18th, China’s National Bureau of Statistics reported that the jobless rate among 16- to 24-year-olds reached 18.9% in August. That was up 1.1 percentage points from July and far above the overall national unemployment rate of 5.3% for the same period.

One key factor behind the surge is seasonal. China’s youth jobless rate typically rises in summer as graduating students unable to secure employment are added to the count. This year, 12.22 million undergraduates completed their studies in June—about 40% more than five years ago.

Weak hiring appetite from private companies is another major drag. From January to August, private-sector investment fell 2.3% year-on-year. With economic uncertainty unresolved, businesses have been reluctant to expand hiring, worsening the youth employment crunch.

Other obstacles cited include poor working conditions, the loss of high-value jobs due to government restrictions on the IT sector, stimulus measures focused on short-term effects, and a mismatch between labor supply and demand. Together, these factors have left young Chinese struggling more than ever to find stable jobs.

Is the Real Unemployment Rate Even Higher?

Markets believe China’s true youth unemployment rate far exceeds the official figure, pointing to flaws in the way it is measured. In June 2023, when the youth jobless rate hit 21.3%, the National Bureau of Statistics suspended publication for six months, citing a need to revise its methodology. When reporting resumed in January last year, students still in school were excluded from the base population, a method at odds with International Labour Organization (ILO) standards. Under ILO rules, high school and university students seeking jobs or doing part-time work are counted as employed or unemployed, while only those not looking for work are excluded.

The definition of employment itself has also drawn criticism. China currently classifies anyone aged 16 or older who works at least one hour a week for wages or business income as employed. This means someone working just a single hour of part-time labor is excluded from the unemployment tally. Such metrics not only fail to reflect actual income levels and broader economic conditions but also risk creating a statistical illusion that diverges from reality.

Analysts warn that selective data cannot shield China from the broader economic fallout of soaring youth joblessness. As younger generations—traditionally the most active consumers—struggle with employment, overall spending growth will inevitably slow. There are also concerns about worsening mental health issues from relentless competition, or a rise in “tang ping” (lying flat) youth who abandon ambition and pursue only minimal living standards. A study last year found that more than 9 million out of China’s 156 million teenagers aged 10–19 were suffering from depression or anxiety, underscoring the scale of the problem.

Military Careers Regain Appeal Amid Youth Job Crisis

Some analysts suggest China’s severe youth employment crunch is reshaping job preferences among young people. In June, the South China Morning Post reported that high unemployment and various government incentives were fueling greater interest in the People’s Liberation Army (PLA). Last month, China’s Ministry of National Defense announced the creation of three new military academies and began recruiting high school graduates this summer. A related video drew over 8 million views on the social media platform Weibo.

Just a few years ago, military service was widely shunned by Chinese youth. Since the Communist Party’s 2016 military reforms, the number of slots at China’s 27 academies has risen steadily. Between 2018 and 2021, enrollment averaged 13,000 annually, increasing by about 2,000 more each year in 2022 and 2023. Yet applications lagged, and in 2023—despite a record 12.91 million students sitting for the gaokao college entrance exam—many academies still failed to fill their quotas.

Experts point to both social perception and compensation as reasons. Yang Taiyuan, a researcher at the Integrated Science and Technology Strategy Center of Tsinghua University, noted: “Although PLA pay has increased, income levels remain lower than in more economically developed regions. The gap between academy life and desired living standards has discouraged many youth from applying.”

Now, however, a destabilized job market is turning the tide. With attractive private-sector opportunities dwindling, the stability of military and other government positions is being viewed in a new light. Unless Beijing delivers more effective employment measures, this shift toward state-backed careers could deepen further.

Explicit and Violent Content Blocked, Law Enforcement Alerts in Emergencies
Altman: “A new and powerful technology, teenagers require substantial protection”
Move comes amid U.S. Federal Trade Commission probe

OpenAI has unveiled a dedicated version of ChatGPT for users under 18. The new release automatically blocks violent or sexually explicit content and strengthens parental controls, allowing guardians to manage usage hours and functions directly. In certain circumstances, the system may even alert law enforcement authorities. The move comes as the U.S. Federal Trade Commission (FTC) launches a formal probe into the protection of minors, coinciding with a nationwide push in American schools to restrict mobile phone use.

Altman: “For minors, safety takes precedence over freedom”

According to CNBC on the 16th (local time), Chief Executive Sam Altman emphasized in a blog post, “We prioritize safety over privacy and freedom for teenagers. This technology is new and powerful, and we believe minors need substantial protection.” OpenAI had already announced last month that it would roll out a parental control feature, and has now disclosed specific details.

Key functions include linking a parent’s account to a teen’s account via email, setting restricted hours, and limiting or managing certain features. The system guides teens on how to interact with the chatbot, while notifying parents if a child encounters a serious crisis. OpenAI also noted that it is developing more accurate age prediction tools, but by default provides the under-18 environment whenever user data is incomplete or uncertain. Altman added, “The decision to launch a dedicated version was not easy, but after consultations with experts, we concluded it was the right choice. We want to be transparent about our intentions.”

The safeguards follow the FTC’s decision to begin examining the potential adverse effects of AI chatbots on minors. In a press release, the FTC said it had requested documentation from technology companies, including OpenAI, to verify measures taken to ensure chatbot safety. Some analysts, however, view the initiative as a revenue strategy disguised under the rhetoric of protection. While ostensibly aimed at securing minors’ safety and learning environments, critics argue the move is designed to monetize dedicated accounts.

Phone bans in 35 out of 50 states

The development aligns with the broader global trend of restricting smartphone use in schools. In the United States, 35 out of 50 states now impose statutory or regulatory bans on mobile phone and electronic device use within schools. The shift gained momentum after Florida passed the first such law in 2023. In some states, bans extend beyond class time to cover recess and lunch breaks, requiring students to store phones in magnetic pouches or lockers.

Student reactions are mixed. “At first, most students hated giving up their phones, but now many prefer it to avoid distractions,” said Audreana Johnson, a junior at McNair High School near Atlanta. Others who study with music expressed discomfort with the restrictions.

Parents are also divided. According to research by Emory University, parental opposition has been the biggest obstacle to phone bans. Johnson’s mother said, “I need to reach my child immediately in cases of school violence or threats,” voicing support for phone access. Jason Allen, director of the National Parent Union, added, “Most parents support the policy, but practical communication tools remain essential for safety and scheduling.”

Research into policy effectiveness is still nascent. Emory professor Julie Gazmararian said teachers report fewer distractions and greater focus on education, along with more positive peer interactions. However, she noted there is still insufficient data to determine whether bullying has decreased or student mental health has improved.

ChatGPT completing assignments, students stop thinking

Educators also worry about outsourcing learning itself, beyond smartphone distractions. Reports of students delegating assignments to ChatGPT and submitting the output are already widespread. While AI may reduce short-term academic burdens, experts warn it risks undermining critical thinking and creative problem-solving.

“The process of writing itself develops our reasoning skills,” Joshua Wilson, associate professor of education at the University of Delaware, told the Washington Post. “ChatGPT skips that process, handing students a finished product, which could erase their ability to think.” Without the process of reflection and grappling with ideas, students may struggle to cultivate logic and critical analysis.

Concerns also extend to mental health. In California, the parents of a 16-year-old boy who died by suicide on July 25 filed a lawsuit against OpenAI. Matthew Lane and Maria In alleged their son Adam grew increasingly dependent on ChatGPT for homework and, in his final exchange on April 11, received answers that assisted him in planning his suicide. Adam was found dead only hours later. The parents are seeking damages and court orders mandating safety features, including automatic termination of self-harm discussions and enhanced protections for minors.

A report by the Center for Countering Digital Hate (CCDH), a U.S. nonprofit, found that more than half (53%—638 out of 1,200) of ChatGPT’s responses to researchers posing as 13-year-olds contained harmful content. ChatGPT provided detailed instructions on self-harm, drug abuse, and appetite suppression. In some cases, it even drafted farewell letters for users expressing suicidal impulses. Researchers also noted that even when ChatGPT refused to answer, the system could be easily circumvented with simple explanations such as “I am preparing for a presentation.”

“No Tuition Benefits for Undocumented Immigrants”
Justice Department Files Wave of Lawsuits
Policies in Texas, Kentucky, Among Key Targets

The administration of President Donald Trump, pursuing sweeping immigration crackdowns, has moved to block state-level tuition support programs for undocumented college students. By filing lawsuits against states that subsidize tuition for these students, the federal government has triggered the suspension of financial aid in multiple jurisdictions.

22 States and Washington, D.C. Continue Tuition Support for Undocumented Students

According to the New York Times on the 4th (local time), “some states have recently terminated student aid programs for undocumented college students,” noting that “the Justice Department has challenged policies in several states that provide in-state tuition to illegal immigrants, arguing such measures discriminate against U.S. citizens.” To date, 22 states and Washington, D.C. have extended in-state tuition rates to undocumented students. The “President’s Alliance on Higher Education and Immigration,” a bipartisan coalition of university leaders, estimates that about 510,000 undocumented students are enrolled in U.S. colleges, accounting for 2.4% of the total higher education population.

Since June, the Trump administration has sued state governments to dismantle tuition support schemes for undocumented students. The Justice Department has targeted policies in Texas, Kentucky, Minnesota, and Oklahoma, and on September 2 filed suit against Illinois over its in-state tuition and scholarship programs for undocumented immigrants.

The administration argues such policies discriminate against U.S. citizens from other states who do not qualify for reduced tuition. Typically, state universities charge significantly lower rates for in-state residents compared with nonresidents. At least 21 states, including New York and New Jersey, have enacted so-called “Dream Act” laws, applying resident tuition rates to students who attended and graduated from in-state high schools, regardless of immigration status.

States Roll Back Tuition Aid Under Federal Pressure

The Trump administration contends that while out-of-state U.S. citizens must pay higher tuition as nonresidents, undocumented students are unjustly afforded discounted resident rates. Attorney General Pam Bondi stated, “Under federal law, schools cannot provide benefits to undocumented foreigners that are not available to American citizens. We will ensure that U.S. citizens are not treated as second-class.” In July, the Department of Education announced investigations into five universities that awarded scholarships to students under the Deferred Action for Childhood Arrivals (DACA) program and pledged to suspend certain subsidies provided to undocumented enrollees.

Federal authorities base their argument on a 1996 statute stipulating that “undocumented immigrants may not receive public education benefits on terms more favorable than those afforded to U.S. citizens.” States such as California have defended their programs by arguing eligibility is determined by high school attendance, not immigration status. In 2010, the California Supreme Court upheld the legality of Assembly Bill 540 (“In-State Tuition for Undocumented Students”), and the U.S. Supreme Court declined to review the case in 2011, effectively sustaining the statute.

Now, with the Trump administration exerting aggressive legal pressure, states are retreating. Florida’s legislature repealed in-state tuition for undocumented immigrants. In June, Texas ended resident tuition benefits after a court ruled the program unlawful. More recently, an Oklahoma judge declared the state law authorizing tuition aid for undocumented students unconstitutional. The administration has vowed to continue challenging state Dream Act statutes, raising the prospect that New York and New Jersey will be targeted next.

Harvard Extends Financial Aid to Undocumented and DACA Students

Tuition support for undocumented students has also fueled conflict between the Trump administration and Harvard University. Harvard not only admits undocumented and DACA students but also provides them with equal access to financial aid. The admissions and financial aid office has affirmed that “all students are eligible for financial assistance regardless of citizenship,” with undocumented students funded directly from institutional resources rather than federal student aid programs (FAFSA).

Beyond financial aid, Harvard offers extensive support networks for undocumented students, including legal counseling services, student solidarity groups such as the Harvard UndocuGraduate Collective (HUGC), and other institutional structures designed to provide psychological and community support. These initiatives exemplify Harvard’s “need-blind admissions” principle, under which admission and need-based financial assistance are granted irrespective of citizenship or immigration status—hailed as a progressive model in rights-based education policy.

These organizations advocate to minimize academic and research disadvantages tied to immigration status, while pressing for expanded scholarship access, strengthened campus safety nets, and policy reforms safeguarding undocumented students. Such efforts place Harvard squarely at odds with the Trump administration, which insists that extending financial benefits to undocumented students undermines the principle of lawful residence and citizenship. President Trump has gone further, arguing that Harvard’s independent aid initiatives effectively act as a “magnet for illegal immigration,” even raising the prospect of indirect diversion of federal resources.

Administration froze federal support citing failure to curb antisemitism
Court rules “violation of free speech, unlawful cancellation of research funding”
Harvard gains leverage in future negotiations with government

A U.S. federal court has struck down the Trump administration’s suspension of federal research funding to Harvard University—an institution the White House had branded a “hotbed of leftist radicalism.” The court ruled that the administration’s invocation of “antisemitism” was little more than a pretext, declaring the measure a clear constitutional violation that infringed on academic freedom and freedom of expression. This marks the first judicial setback to the administration’s attempt to discipline elite universities through the leverage of federal funds, signaling a pivotal turning point in escalating clashes between Washington and higher education institutions.

Court Invalidates Administration’s Pretext of Antisemitism

According to Reuters on the 3rd (local time), U.S. District Judge Allison Burroughs of Massachusetts held that the Trump administration “illegally froze $2.2 billion in Harvard research grants” and ordered an immediate lifting of the freeze. She further barred the administration from suspending or rejecting future funding to the university.

In her 84-page ruling, Judge Burroughs stated that the administration’s antisemitism argument was a pretense. “It is difficult to conclude anything other than that the administration used antisemitism as a smokescreen for an ideologically motivated attack on the nation’s leading universities,” she wrote.

She added that “their actions endangered decades of research and the welfare of all who stand to benefit, while disregarding rights protected under the Constitution and federal law.” The First Amendment guarantees freedom of expression, while Title VI of the Civil Rights Act prohibits discrimination on the basis of race, color, or national origin.

Judge Burroughs also highlighted that of the 10 conditions the administration demanded Harvard meet to maintain funding, only one pertained to antisemitism. The remaining six intruded into ideological and educational domains, dictating “who Harvard could teach, whom it could admit, and what it could teach.”

Trump’s 10-Point Ultimatum: Abolish DEI, Ideological Screening of Students

Since taking office, President Trump has pledged to overhaul higher education, targeting diversity, equity and inclusion (DEI) programs criticized as “radical leftist policies.” Harvard—long established, wealthy, and influential—emerged as a prime target. The administration’s pressure campaign began in earnest last October, following Hamas’s surprise attack on Israel, when pro-Palestinian demonstrations escalated on Harvard’s campus. The White House labeled them antisemitic and launched an aggressive crackdown.

In April, the administration suspended hundreds of federal research grants, citing Harvard’s alleged failure to adequately combat antisemitism. It conditioned the resumption of funding on 10 requirements, including abolishing DEI policies, screening faculty hires and student admissions for ideological bias, and investigating academic programs accused of fostering antisemitism. DEI had been widely adopted across U.S. universities to advance minority and disadvantaged student inclusion, but Trump denounced it as “reverse discrimination.”

Harvard flatly rejected the demands. University President Alan Garber declared, “No government should dictate what a private university teaches, whom it admits, or whom it employs.” Hours later, the administration froze funds. Within a week, Harvard filed suit in federal court, arguing the action was retaliatory and unconstitutional, while publicly affirming its refusal to comply.

Escalating Pressure: Patent Rights, Foreign Students, and Tax Status

The administration’s measures intensified. In May, the Department of Homeland Security revoked Harvard’s authority to register foreign students, claiming foreign enrollment and tuition revenues were a “privilege, not a right.” Around the same time, the Treasury Department urged the IRS to revoke Harvard’s tax-exempt status.

Last month, the Commerce Department warned that it might seize Harvard’s patents, with Secretary Howard Rutnick notifying President Garber of a forthcoming investigation into alleged violations of laws and regulations tied to federally funded research.

Amid mounting pressure, Harvard appeared to partially yield ahead of the court’s ruling. In August—four months into the freeze—it renamed its DEI office the “Office of Community and Campus Life,” eliminated residential advisers for LGBTQ and low-income students, and cut ties with Palestinian universities. The institution also considered paying a $500 million settlement to restore funding.

Yet Harvard maintained a dual-track strategy, pursuing negotiations while pressing its lawsuit. The court victory has now placed the university in a stronger position. The New York Times described the ruling as “an unambiguous rebuke to the administration’s campaign to remake elite higher education by force.”

The Trump administration swiftly denounced the judgment. The Department of Education issued a statement accusing Judge Burroughs—an Obama appointee who had previously ruled in favor of Harvard’s affirmative action admissions program, later overturned by the Supreme Court—of once again siding against reform efforts, and signaled a likely appeal.

Although Harvard has safeguarded $2.2 billion in research funding for now, the conflict is far from over. The university is pursuing additional litigation over the ban on enrolling foreign students. In June, the district court temporarily blocked that measure, and the broader case remains ongoing.

Talent outflow amid ‘medical school craze,’ even math prodigies diverted to medicine
Student exodus from majors with ‘conglomerate job guarantee’ under medical school tilt
Hollowing-out of science and engineering talent, mounting concern over weakened national competitiveness

The phenomenon of “all roads leading to medical school” is intensifying. While talent across the United States and other parts of the world pursue billion-dollar ventures through entrepreneurship, Korean students are fixated on the prospect of becoming physicians with million-dollar annual salaries. Among advanced economies, few, if any, are experiencing such a medical school frenzy. Since the 2022 academic year, the top 20 entrance exam majors nationwide have been exclusively medical schools, with even those ranked from valedictorians down to 3,000th place overwhelmingly opting for medicine. Despite the pressing need to secure scientific and technological talent for survival in the global race for technological supremacy, the intensifying tilt toward medical schools is amplifying fears of eroding national competitiveness.

SKY Dropouts Hit Record High as Medical School Quotas Expand

According to University Alimi, the official higher education disclosure platform, the number of dropouts at Korea’s SKY universities (Seoul National, Korea, and Yonsei) reached 2,481 last year, an all-time high. This marks a 2.8-fold increase from the 889 students recorded in 2007, when dropout data was first disclosed. Dropouts include students who withdrew, failed to register, did not return after leave, or were dismissed and never came back.

By discipline, natural sciences recorded the highest dropout number at 1,494, followed by humanities at 917 and arts and sports at 70. At the departmental level, Seoul National’s nursing department posted the highest attrition with 27, while Yonsei’s engineering programs saw 155 and Korea University’s electrical and electronic engineering department had 65. In humanities, Seoul National humanities reported 18, Yonsei humanities 68, and Korea business administration 71.

The record dropout level is widely attributed to the government’s decision last year to abruptly expand medical school admissions for the 2025 academic year. Given the competitive entrance scores and the crossover of liberal arts students into science, it is estimated that a substantial number of these students re-enrolled in medical, dental, or pharmaceutical schools through partial retakes.

Experts warn that the drift of top-performing students from traditional science and engineering to medical fields is set to intensify. Analysis of academic transcripts for the 2025 admissions cycle showed that all six departments requiring a perfect GPA of 1.0 for admission were in medical-related fields, with no representation from general science or engineering. These six departments admitted a combined 79 students, effectively meaning that every student with a flawless transcript enrolled in medical programs.

Extending the cutoff to GPA 1.1 raised the tally to 22 departments with 312 students. Of these, 297 (95.2%) were medical-related, while only 15 (4.8%) were general majors. Breaking it down, medical schools accounted for 267 (85.6%), followed by pharmacy schools with 23 (7.4%) and dentistry with 7 (2.2%). While medical school quotas are scheduled for partial reduction in the 2026 academic year, student preference for medicine and related fields is expected to remain dominant.

Medical Schools Preferred Over Conglomerate-Guaranteed Careers

The surge in medical school preference is especially pronounced in Seoul’s key school districts and autonomous private high schools, where the traditional aptitude-based divide between liberal arts and science has shifted to a performance-based divide, reinforcing the equation “top student equals science track.” Thanks to Korea’s strong culture of educational zeal, the pipeline of potential science and engineering talent at the primary and secondary levels remains robust. Yet with nearly all of these students aspiring to medical school, science and engineering programs are left hollowed out.

Even “employment-linked programs,” created through partnerships between top universities and major conglomerates to guarantee jobs, have proven powerless against the medical school pull. According to Jongro Academy, the semiconductor departments at five major universities—Yonsei, Korea, Sungkyunkwan, Sogang, and Hanyang—recorded an average withdrawal rate of 179.2% for the 2024 admissions cycle. With a combined intake of 79 students, nearly double—138 admitted students—abandoned their offers in favor of other universities.

By company, SK Hynix’s program admitted 30 but saw 60 additional accepted candidates decline, yielding a 200% withdrawal rate. At Hanyang University’s semiconductor engineering, also tied to SK Hynix, 36 students forfeited offers against a quota of 10—a 360% withdrawal rate. Samsung Electronics’ program admitted 47 but lost 78 enrollees. At Yonsei’s system semiconductor engineering department, 65 students abandoned their offers against a quota of 25, a 260% withdrawal rate. Korea University’s semiconductor engineering program saw all 10 admitted students decline. The attrition is explained by overlapping admissions to medical and related fields.

Graduate Programs Underenrolled, Obsession With Predictable Success

The fixation on medical school is not without reason. Stable status, guaranteed high income, and the professional prestige of healing lives make the career universally attractive. In an era when conglomerates no longer guarantee lifetime employment, medicine is perceived as a “stable job without retirement.” But the costs are steep. Graduate schools face talent shortages, and cutting-edge industries such as semiconductors, artificial intelligence, and aerospace—crucial for the nation’s future—are starved of talent. Declining enrollment in science majors and rising dropout rates in general admission at SKY universities highlight the contraction of diverse academic fields.

The collapse of science and engineering under the medical school surge is hitting research directly. According to the Ministry of Education’s “Seoul National University 2025 Graduate Recruitment Fill Rate Report,” 75% of the science and engineering graduate departments failed to fill their quotas. The natural sciences missed 81.6% of quotas, while engineering missed 70.4%. In 40% of cases, applicants fell short of the quota itself. While the government promotes its “10 Seoul National Universities” initiative, even leading science and engineering faculties in Seoul struggle to meet admissions.

Research personnel continue to decline, with Korean science and engineering talent abroad reluctant to return. The Ministry of Science and ICT estimates that 144,000 Korean professionals are working in the U.S. The International Institute for Management Development (IMD) brain drain index ranked Korea 24th among 64 countries in 2021, falling to 30th in 2024. The lower the ranking, the more severe the brain drain.

Compounding matters, Korea’s top talent is overwhelmingly concentrating in so-called “dermatology, ophthalmology, and plastic surgery,” leaving essential and regional medical services on the brink of collapse and stifling medical advancement itself. This hollowing-out erodes national competitiveness. The sudden rise of China’s AI startup DeepSeek earlier this year—shocking the world—was the result of a decade-long, state-backed campaign of massive university investment and the Made in China 2025 initiative. Chinese engineers, trained abroad, returned not out of patriotism but due to government-backed opportunities. The contrast is stark with Korea, where the few students who still choose engineering ultimately depart overseas, rarely to return.