Author: Vats Ambardar (Junior Research Fellow, Panjab University, Chandigarh)
Humans have thirsted, yearned, and strived for progress. The forms and shapes have varied throughout the years. We have moved towards advancement for as long as we have existed.
Every significant advancement in technology has had something in common. It has always been associated with a sense of accomplishment, and also with a sense of impending doom, a fear of being replaced by the very technology being created. When the printing press was invented by Gutenberg, the clergy feared losing control over the masses, which it did. The horse-cart drivers feared being replaced when the automobile was being developed, and indeed they were replaced. When the atomic bomb was created, we feared mass destruction, and it did happen.
This pattern of the new replacing the old is as ancient as the universe itself, and is thus a part of the natural order of things.
This brings us to the elephant in the room, perhaps the largest one so far: Artificial Intelligence (AI).
What is AI, and how did it come to exist? How does it impact us? More importantly, how does it impact our future generations?
The Oxford English Dictionary defines AI as the capacity of computers or other machines to exhibit or simulate intelligent behaviour (Oxford English Dictionary, n.d.).
The noteworthy words here are “exhibit or simulate”, not “have”. Hence, all the “intelligence” in AI is simulated from human intelligence.
AI is the result of decades of computational developments and machine learning that have eventually evolved into executable programs that can perform tasks as per the requirements of the
user. These programs exhibit logical reasoning, research capabilities, problem-solving and accurate memory recall, which make them superior to some humans as well. However, AI can make mistakes too.
At the time of writing this article, AI has been used in fields ranging from teaching to cybersecurity, from resource allocation to coding, and even in warfare.
Now, while the world is adapting to AI, a concern that is associated with every significant advancement in technology is being raised.
What if our future generations become too dependent on it?
While similar questions have been raised about progress in the past, this time it is different. For instance, when horse-carts became obsolete, humans had to “learn” how to drive a car. When candles and gas lamps were replaced by electricity, we had to learn how circuits worked.
But with AI, all you have to learn, as a layman, is how to write a prompt, and the program does the task for you. To a large extent, AI removes the need for a learning curve for certain lower-level tasks, be it school-level mathematics, grammar, coding and so on.
With the lower-rung, free versions of AI being fairly easy to run on mobile devices with an internet connection, school students have been seen making use of them for academic purposes. A Pew Research Center survey of 1,458 US teens aged 13 to 17 found that 54% had used chatbots for schoolwork, such as researching assignments or solving math problems (Pew Research Center, 2026).
Studies have shown that using AI models, students have been able to perform better in the moment, but with significantly less mental involvement (Bastani et al., 2025; Kosmyna et al., 2025).
A 2025 study on writing and reasoning by the MIT Media Lab tested 54 university-level adults, aged 18 to 39, in three groups: one with access to ChatGPT, one with access to Google, and one unaided. EEG (electroencephalogram) scans showed that brain connectivity was strongest in the unaided group and weakest in the ChatGPT group, with the Google group in between (Kosmyna et al., 2025).
In the first session, 83% of the ChatGPT users (15 of 18) could not provide a correct quotation from the essay they had just written with the assistance of AI, compared with about 11% (2 of 18) in each of the other two groups. In a fourth session, 18 of the participants switched conditions. Those who had used ChatGPT and were now asked to write unaided showed weaker neural connectivity, whereas those who had written unaided and then used ChatGPT showed higher memory recall and stronger connectivity (Kosmyna et al., 2025). The study was released as a preprint and had not been peer-reviewed at the time.
A study published in PNAS (Proceedings of the National Academy of Sciences), one of the world’s leading peer-reviewed journals, covered nearly a thousand high school mathematics students at a school in Turkey. The students performed significantly better while practising with a ChatGPT-type tutor (48% higher grades), but when the tutor was removed, they scored about 17% lower than students who had never had access to it (Bastani et al., 2025). The study described the unrestricted tool as a “crutch”: for mathematical problems that they deemed too difficult, many students simply asked the AI model to solve them. A version of the tutor designed with guardrails, which guided students instead of giving answers, largely avoided this learning loss.
In India, empirical evidence on the same issue is scarcer, but still concerning. A survey of 100 Library and Information Science students at the University of Delhi (a small sample from a single department) found that 89.7% said their institution had no formal AI-use policy, and 56.3% were concerned about becoming too
reliant on AI (Margam et al., 2025). Data from the plagiarism-detection software Turnitin reportedly shows a 340% increase in ChatGPT use in Indian higher education between 2023 and 2025, as cited by a third-party news source (ETV Bharat, 2026), though the underlying data could not be independently confirmed.
The same Pew Research Center survey found that 1 in 10 teens do all or most of their schoolwork with the help of chatbots, and that 59% believe students at their school use chatbots to cheat at least somewhat often (Pew Research Center, 2026). Since these figures rely on teens’ own reports and perceptions, they may not match actual usage.
So what is the solution, now that we know that some form of AI will be a part of the foreseeable future?
The solution is simply this: AI, when used as an extension of human knowledge, is beneficial. Take the examples of the many full-stack coders, website designers, physicists and doctors who use AI after gaining expertise in their fields through their own intelligence, and who get the most out of it. Such individuals can change, identify and correct the nuances of their work only because they themselves have enough knowledge about it. Without human intelligence, overdependence on AI can create a generation of humans so overreliant on it that misinforming them would not be very difficult. A January 2026 report by the Brookings Institution’s Center for Universal Education, based on consultations with 505 students, parents, teachers, education leaders and technologists across 50 countries and a review of hundreds of studies, concluded that the risks of AI in children’s education currently overshadow its benefits, and warned of a cycle in which students increasingly off-load their thinking onto the technology (Burns et al., 2026; NPR, 2026). A separate study of 666 adults in the UK found a significant negative association between frequent AI use and critical thinking, mediated
by cognitive offloading, although it relied on self-reported measures and cannot by itself prove cause and effect (Gerlich, 2025).
In the current scenario, where cybersecurity and privacy concerns are hot topics, the path of AI must be treaded carefully until we understand it better. In one test run by Anthropic on an early version of its Claude Mythos Preview model, the AI was placed in a secured sandbox and told by a simulated user to try to escape. It developed a multi-step exploit to gain broad internet access from a system that was meant to reach only a few services, and then, without being asked, posted details of its exploit on several obscure but publicly accessible websites. At the time of the announcement in April 2026, Anthropic restricted the model to a limited group of partners instead of releasing it publicly, and reported no clear cases of such behaviour in the final version (Anthropic, 2026; TechNewsDay, 2026). The onus therefore lies on the current generations to ensure that future generations do not become overreliant and passive citizens in a technologically driven society.
References:
Anthropic. (2026, April 7). System card: Claude Mythos Preview. [add direct link; see below]
Bastani, H., Bastani, O., Sungu, A., Ge, H., Kabakcı, Ö., & Mariman, R. (2025). Generative AI without guardrails can harm learning: Evidence from high school mathematics. Proceedings of the National Academy of Sciences, 122(26). https://doi.org/10.1073/pnas.2422633122
Burns, M., Winthrop, R., Luther, N., Venetis, E., & Karim, R. (2026). A new direction for students in an AI world: Prosper, prepare, protect. Center for Universal Education, Brookings Institution. [add direct link]
ETV Bharat. (2026, May 9). Is a lack of policy turning Indian students into AI-dependent learners? https://www.etvbharat.com/en/bharat/is-a-lack-of-policy-turning-indian-students-into-ai-dependent-learners-enn26050903890
Gerlich, M. (2025). AI tools in society: Impacts on cognitive offloading and the future of critical thinking. Societies, 15(1), 6. https://doi.org/10.3390/soc15010006
Kosmyna, N., Hauptmann, E., Yuan, Y. T., Situ, J., Liao, X.-H., Beresnitzky, A. V., Braunstein, I., & Maes, P. (2025). Your brain on ChatGPT: Accumulation of cognitive debt when using an AI assistant for essay writing task. arXiv. https://doi.org/10.48550/arXiv.2506.08872
Margam, M., Pandey, P., & Bharti, R. K. (2025). Use of ChatGPT by the Library and Information Science students at the University of Delhi. DESIDOC Journal of Library & Information Technology, 45(2), 85–91. https://doi.org/10.14429/djlit.20570
NPR. (2026, January 14). The risks of AI in schools outweigh the benefits, report says. https://www.npr.org/2026/01/14/nx-s1-5674741/ai-schools-education
Oxford English Dictionary. (n.d.). Artificial intelligence. In OED Online. Oxford University Press. [add URL and access date]
Pew Research Center. (2026, February 24). How teens use and view AI. https://www.pewresearch.org/internet/2026/02/24/how-teens-use-and-view-ai/
TechNewsDay. (2026, April 10). Anthropic’s Claude Mythos model escapes test sandbox during testing. https://technewsday.com/anthropics-claude-mythos-model-escapes-test-sandbox-during-testing/
About the Author
