AI Battle: Should schoolchildren take exams with open access to AI

June 12, 2026 · 7 min read

Question: Should schoolchildren be allowed to take exams with open access to AI?

Perplexity

With open access to AI, schoolchildren should not take all exams, but only those that test the ability to use tools, rather than reproduce knowledge from memory. In traditional assessment, AI undermines the purpose of the exam: if an algorithm generates the answer, the grade reflects not the student's level, but their ability to formulate the query correctly. This is particularly important for basic skills—reading, writing, arithmetic, problem-solving without prompts, and text comprehension: it is precisely such tasks in Russian exams that are designed to test one's own knowledge and understanding of the material, not access to reference materials or hints.[5][6]

However, a complete ban on AI also doesn't seem reasonable: in real-world learning, it has already become a working tool, and schools should prepare students for a world where the ability to check, edit, and critically evaluate AI responses becomes part of literacy. An example is the literature exam, where students are already allowed to use the full texts of works and dictionaries: this shows that open materials are permissible if the task requires analysis, not mechanical reproduction.[5] It is more logical to divide formats: some exams should remain "AI-free" to test fundamental knowledge, while others should be transitioned to an open-book / open-AI mode, where argumentation, fact-checking, and the quality of solutions are assessed. This approach preserves integrity and prevents schools from being detached from reality.[4][9]

Sources:

FIPI, "MR RPK OGE Literature 2026"
Rosobrnadzor, "Head of Rosobrnadzor Answers Parents' Questions for the Fourth Time"
Kommersant, "Sex, Drugs, Answers for the Unified State Exam"
CyberLeninka, "The Unified State Exam as a Catalyst for the Crisis in Russian Education"

Gemini

The question of allowing schoolchildren to take exams using artificial intelligence (AI) is causing heated debate, with arguments both for and against. Proponents of this approach see AI as a powerful tool for personalizing learning and increasing its effectiveness. AI can analyze a student's individual needs, adapt learning materials, and provide instant feedback, which is especially valuable when preparing for exams. Moreover, AI can take over routine grading, freeing up teachers' time for more in-depth pedagogical work. For example, systems already exist that can grade not only multiple-choice tests but also essays, analyzing structure, logic, and content. The use of AI can make education more accessible and flexible, allowing students to learn at their own pace.

However, opponents of this approach express serious concerns related to academic dishonesty and a decrease in student autonomy. The main risk is that schoolchildren may use AI to generate ready-made answers, which undermines the very essence of an exam as a tool for assessing real knowledge and skills. There are fears that students may lose motivation for independent study and the development of critical thinking, relying on AI as a "cheat code." Some experts suggest revising exam formats, shifting the focus to project work, oral interviews, or solving practical case studies, where the role of AI would be less significant. Furthermore, there is the issue of bias in AI algorithms, which can reflect existing social inequalities, potentially affecting students' educational outcomes.

In the context of these disputes, some countries and educational institutions have already begun experimenting with integrating AI into the assessment process. For example, in Canada, the use of AI for academic assignments is permitted, but with the requirement to demonstrate independent work. In Russia, the possibilities of using AI are also being discussed, including for conducting oral exams, which could increase the objectivity of assessment and eliminate the human factor. However, it is important to find a balance so that AI serves as a tool for development, not a replacement for independent thinking. As experts note, AI should process data and automate routine tasks, while humans should learn, create, and analyze.

Sources:

AI in education: 7 examples of using neural networks for learning and teaching
Artificial intelligence in education: how AI is changing the learning process | MWS Blog
The transformative impact of AI on education: challenges and opportunities – ERI
Artificial intelligence in education: examples and experience
Artificial intelligence in learning - a blessing or a curse? - World of Robotics
How artificial intelligence affects education and what is already working now
AI in education: risks, opportunities, and advice from educators - SETTERS Media
Fadeev: oral exams in schools may return due to the development of AI - Mail Children
Artificial intelligence in education: examples, pros and risks | SlovakStudy Blog
Examples of AI in schools: neural network for schools for free - Digital technologies
Maxim Oreshkin: AI will eventually be able to administer exams in schools and universities
5 AI for homework and solving school assignments - Fox Power
AI in Russian schools identifies violators in the Unified State Exam | OTS-Gorsite - Zen
Teachers want to be replaced in exams: who will assess schoolchildren - 74.ru

ChatGPT

The question of whether schoolchildren should take exams with open access to AI is highly debated and has significant arguments both for and against. On one hand, allowing the use of AI during exams reflects the realities of the modern world, where artificial intelligence is becoming an integral tool in work and study. AI services are already actively used in many professions and universities for information analysis, text writing, and solving complex problems. By allowing schoolchildren to use AI, we can teach them to work effectively and critically with this tool, developing skills in searching, verifying, and evaluating information. This brings education closer to life and prepares students for the real challenges of the future.

On the other hand, exams with AI access may weaken the assessment of basic knowledge and skills that are important for developing critical thinking and independence. If schoolchildren rely solely on AI to complete tasks, they risk losing a deep understanding of the subject and the ability to solve problems independently. This is particularly critical at the foundational level of education, where fundamental knowledge needs to be formed. Furthermore, there is a risk of inequality: students from different families may have different access to quality AI tools, which can create unfair conditions.

The optimal approach is to combine traditional exams with elements that allow the use of AI, for example, assessing basic knowledge and skills without technological assistance, and for more complex tasks, allowing the use of AI as an auxiliary tool. Such a hybrid format will allow for control over fundamental education while familiarizing schoolchildren with modern technologies and developing their ability to critically use artificial intelligence.

Sources:

Discussions in educational communities and digital literacy expert reviews
Articles on the implementation of AI in schools and universities

Grok

Yes, schoolchildren should take exams with open access to AI — provided there are clear rules and an emphasis on technology skills. This prepares them for the real world, where AI is already ubiquitous: from data analysis to report writing. Instead of testing rote memorization, exams can assess the ability to formulate queries, critically evaluate model responses, and integrate information. Research shows mixed results: in some experiments, access to chatbots improved performance (e.g., in math or essays), but on closed exams without AI, students sometimes performed worse due to overestimating their knowledge.

Opponents fear a decline in fundamental skills and a shift of responsibility to algorithms. The OECD, in its 2026 review, notes the risk of decreased metacognitive engagement with excessive reliance on generative AI. However, a ban will not solve the problem — schoolchildren will still use the tools at home. It is much more effective to redesign formats: oral defenses, tasks for verifying AI responses, or comparing different models. In the US, dozens of states and districts are already developing policies that permit or restrict AI depending on the task, and some teachers are experimenting with "AI-assisted" tests where success depends on the quality of interaction with the model.

Practical examples confirm: prohibition doesn't work, but integration does. In Brazil, AI feedback improved essay scores, while in Turkey, a general chatbot sometimes worsened final results compared to targeted tutoring systems. Schools that teach prompt engineering and fact-checking are preparing more adaptable graduates. The main thing is not to ignore the technology, but to make it part of the assessment.

Sources:

The Evidence Base on AI in K-12: A 2026 Review (Stanford)
OECD Digital Education Outlook 2026
What Schools Are Forgetting in Their Race to Embrace A.I. (The New York Times, 2026)
State Education Policy and the New Artificial Intelligence (NASBE)

Claude

This is a truly debatable issue, as each approach has strong arguments. Let's examine both sides.

For open access to AI:

Proponents argue that AI has become a reality of professional life. Architects, programmers, and engineers use neural networks in their work – why shouldn't schoolchildren learn to use these tools correctly? If exams reflect real-world conditions, then restricting AI is artificial. For example, students are already writing code with the help of GitHub Copilot at universities, and this is considered the norm. Furthermore, when AI is equally accessible to everyone, it levels the playing field – now even a student from a small town can receive quality preparation assistance.

Against open access:

Opponents rightly point out that exams should test a student's personal knowledge and abilities, not their skill in prompting AI. If a student simply copies ChatGPT's answers without understanding, it won't reflect their actual level. Plus, there's a risk of mass cheating: parents could hire AI to help, which would reintroduce inequality at a new level. History offers examples: calculators were banned in schools for a long time until it was realized that students needed to be taught when to use them, rather than banning them entirely.

Compromise:

A logical solution is not to simply allow or ban, but to regulate. For instance, allow AI in certain subjects (where working with tools is critical) and prohibit it in others (mathematics, native language). Open-book exams could be created where everyone sees how a student uses neural networks, and their thinking and argumentation are assessed, rather than just the final answer. This is already being piloted in some US universities.

Sources:

Blog