TOPIK exam being sold to Naver: Answers graded by AI, Paper-based tests discontinued
Korea's premier language exam is undergoing its biggest transformation since the first edition back in 1997. A consortium led by tech giant Naver is set to acquire operational rights to the TOPIK (Test of Proficiency in Korean) through a ₩350 billion deal, putting the government-administered test into private hands.
The sweeping changes include a complete elimination of paper-based testing by 2026, test fees increasing up to 75%, and artificial intelligence replacing human graders. Many educators are seriously concerned about accessibility and fairness for the nearly half million annual test-takers worldwide.
Let's look at the full impact of TOPIK's privatization, from digital transformation, AI assessment, fee increases, accessibility concerns, and mounting professional opposition.
TL;DR
- Naver is set to pay ₩350 billion for TOPIK's rights: from test creation and scoring to administration and revenue generation
- Grading of answers and creation of exam questions to be done by AI
- Traditional paper-based tests will be phased out, replaced by digital-only formats
- Test fees will increase up to 75%
- Over 11,000 language professionals have opposed the move
- Naver's incentives to showcase its own AI could compromise grading accuracy
TOPIK's importance
Nearly 500,000 people worldwide took TOPIK in 2023. Like China's HSK for Mandarin or Japan's JLPT for Japanese, TOPIK is the definitive measure of Korean language skills across multiple contexts. However, unlike TOPIK's new for-profit model, these and other major global exams are run by government bodies or non-profit foundations.
For job seekers, TOPIK scores often determine employment opportunities, with many Korean companies using TOPIK 4, 5 or 6 as a hard requirement. Foreign workers and immigrants also rely on these scores to meet visa requirements and secure residency status, giving the exam an official role that diretly affects the ability of many to build up a long-term future in Korea.
In higher education, TOPIK functions as both an entry and exit requirement. International students often need qualifying scores for university admission, while many must also meet TOPIK benchmarks to graduate. The exam similarly influences scholarship opportunities, with many programs requiring specific levels.
This role has been built on decades of government oversight and standardized administration. Now, with these fundamental changes looming, that role faces its greatest test.
The Digital Transformation
This year (2025), TOPIK has started introducing digital testing (IBT) alongside its traditional paper format. By 2026, test-takers will lose the option of paper-based exams entirely.
The new AI-powered system represents an even more dramatic departure from tradition. Rather than limiting artificial intelligence to multiple-choice questions, the system will evaluate complex writing responses—areas that have traditionally required human judgment to assess cultural nuance and linguistic subtlety.
Education officials promise faster results and consistent grading standards, but critics point to a troubling gap: the absence of comprehensive validation studies for AI-based Korean language assessment. As one professor noted:
This is a matter that has never been verified even once through objective methods such as academic papers regarding the validity and reliability of AI-based TOPIK question generation and automatic scoring that the Naver consortium intends to introduce.
Testing Fee Hikes
The financial burden on test-takers will be substantial. Current fees of ₩40,000 for TOPIK I and ₩55,000 for TOPIK II will jump to ₩70,000 and ₩95,000 respectively during the initial rollout phase (2026-2030). While officials promise reductions after 2031 to ₩55,000 and ₩75,000, even those "reduced" rates will exceed current pricing.
These fee hikes are particularly problematic for test-takers from developing countries, who may need to retake the exam multiple times to achieve required scores for their specific goals. The irony isn't lost on critics: despite promises of technological efficiency and increased testing frequency, test-takers will pay significantly more, not less.
Digital Divide Challenges
The mandatory shift to digital-only testing creates barriers unrelated to Korean language ability. As one university professor pointed out, "Korean language learners who are just starting to learn Korean find it difficult to write Hangeul by hand, but struggle even more with typing Korean on keyboards. This change lacks proper consideration of these aspects."
The planned "home testing" option, while offering flexibility, raises additional equity concerns. Not everyone has access to quiet, private testing environments with stable internet connections, potentially creating uneven conditions that could affect scores and, by extension, professional and academic opportunities.
Professional Pushback
The academic and professional community has mounted significant opposition. As of May, at least 11,025 Korean language educators and researchers have signed a petition opposing the privatization, citing concerns ranging from testing integrity to fundamental questions about access and fairness.
Lee Chang-yong, head of the Korean Language Teachers' Branch at the Workplace Gapjil 119 Online Union, emphasized the public nature of TOPIK:
TOPIK is a public good that serves as an entry point for foreigners and immigrants to enter Korean society and as a road to navigate within the community. The privatization attempt under the guise of digital transformation of the Korean language proficiency test is fundamentally shaking the essence of Korean language education.
The absence of meaningful consultation has intensified opposition. Lee noted that "this project completely lacked a public discussion process that transparently collected diverse opinions from stakeholders such as Korean language education experts and learners. The fact that there was not even a public hearing, which is essential in the policy promotion process, proves that this policy was pushed unilaterally and unreasonably."
One university professor called the privatization "as absurd as calling the 수능 college entrance exam 'private profit-making software'", criticizing the government for "viewing Korean language education through market logic and having no awareness of the problem that Korean as a second language should naturally be provided as educational infrastructure."
The Naver Consortium Deal
Education ministry officials describe the arrangement as a "public-private partnership," but in reality the private role appears very dominant. The Naver-led consortium, which includes NS Develop and Daekyo, gains operational control over virtually every aspect of TOPIK: from test creation and scoring to administration and revenue generation.
Beyond test operations, the consortium has secured rights to develop and market supplementary learning materials, effectively controlling both the assessment tool and preparation resources. This vertical integration raises questions about conflicts of interest and monopolization.
The consortium will also gain access to valuable data from hundreds of thousands of annual test-takers worldwide, information that could prove commercially valuable beyond the immediate scope of language assessment.
Conflict of Interest in AI Model Selection
Naver's investment in developing proprietary AI models creates an inherent conflict of interest in TOPIK's digital transformation. The company has strong incentives to showcase its own models as a "success story" and gain valuable experience with large-scale deployment, regardless of whether these models deliver optimal results for test-takers.
This approach would contradict what should be the primary goal: selecting the most accurate models without bias, on a task-by-task basis. Through our development of TOPIK Easy6, we've discovered that no single AI model excels at all Korean language tasks. When we test different models across various assessment components, each shows distinct strengths and weaknesses, making it clear that optimal testing requires careful model selection for each specific evaluation task.
Our testing has also revealed that accuracy comes at a price - literally. We've found that only the most expensive, cutting-edge models consistently deliver the accuracy levels necessary for fair assessment. Yet we've made another troubling discovery: a model's ability to produce natural-sounding Korean, demonstrate cultural knowledge, or perform translation tasks shows surprisingly little correlation with its accuracy in TOPIK-specific functions like grading, proofreading, or providing meaningful feedback. This disconnect between general Korean proficiency and suitability for TOPIK grading and teaching has been one of our biggest challenges that we had to overcome to build a reliable preparation tool.
These technical aspects suggest that Naver's likely preference for its own models, driven by business imperatives rather than accuracy, could systematically disadvantage test-takers. Instead of receiving evaluations from the most capable AI for each specific task, students may face assessment by models chosen primarily to advance Naver's commercial AI ambitions.
Timeline and Implementation
The transformation will unfold across several phases. The 2025 hybrid model will maintain paper options while introducing digital alternatives. By 2026, paper-based testing disappears entirely, though testing frequency will double from six to twelve annual sessions.
Between 2029 and 2030, testing opportunities will double again to 24 annual sessions, while AI-proctored "home testing" becomes available. The final phase, launching in 2031, promises "on-demand" testing that allows individuals to schedule exams at their convenience.
Testing the Testers
TOPIK's privatization ultimately tests something more fundamental than Korean proficiency: Korea's commitment to accessible pathways for global participation in its society.
When academic critics compare this move to privatizing the 수능, they're highlighting an essential contradiction. Korea maintains fierce public control over tests that affect Korean citizens while commodifying the test that affects everyone else. The message, intended or not, is clear about who is listened to and who is not even asked for their opinion.
The real evaluation won't come from AI algorithms or quarterly earnings reports. It will come from the Burmese student who can no longer afford multiple test attempts or the Pakistani engineer whose rural internet connection fails during home testing. These individual stories will accumulate into a collective verdict on whether Korea chose corporate efficiency over public responsibility.