At our core, we believe Deskpad is a mission. We are a group of high school students, and are concerned about AI’s impact on education. As profit-motivated AI labs push out better and more sophisticated frontier models, the industries that are slow to adapt have been the ones impacted most deeply. Education, which is built on the core values of transparency and accountability, has been hit particularly hard by AI’s continuous evolution. We have witnessed firsthand how our fellow classmates have explicitly taken assignments, uploaded them into ChatGPT, and copied-and-pasted the output back into their documents. This cheating has gotten so bad that schools across the country have had no choice but to ban the use of AI and technology altogether. While I once looked at many of these educators in disdain, after speaking to more than one hundred in my development of Deskpad, I have come to develop a deepening sense of sympathy for the seemingly insurmountable challenges they are currently facing in their classrooms.

What we created Deskpad to do, fundamentally, was to provide an alternative to schools who felt the need to ban AI in the classroom. Deskpad is not anti-AI. We are techno-optimist—of the belief that AI has the potential to do tremendous good for our society, and today’s students will be the ones pioneering that change. But all of these benefits will only exist if AI is adopted, and more importantly, adopted in the right way. Schools, to this point, have taken the wrong approach to AI. We believe that the solution to AI’s incorporation education is a guardrailed, transparent approach with both students’ and teachers’ best interests kept at heart.

The risk with AI in education, of course, is also especially large. The technology that allows a student to learn at a pace never seen before also has the potential to make them incredibly stupid. There is published research that an over-reliance on AI, especially in an academic setting, can erode the part of your brain that thinks critically. If AI is to be adopted in education, there are also immense concerns around AI’s environmental impact, student data privacy, and model hallucination on important facts. We believe, however, that several of these risks are mitigated when thoughtful and student-centered adoption is a conscious choice by schools. I will address our perspective on each of the previous concerns below:

1. On cognitive decline: If every AI model incorporated into the classroom is transparent and guardrailed, this becomes a significantly lower risk. Students who use AI as a “thought partner” are freed to produce their best work while still owning all of the content and ideas. While there will still be a need for some level of student surveillance (as there has been across education’s history), in a world where AI’s usage is allowed thoughtfully, student cheating will decline. The core idea is that when the technology is encouraged with the student’s best interests at heart, students will inherently begin to understand where AI will help or hurt them. In a world where the technology is flat-out prohibited, students will look for loopholes and workarounds, which is where the cognitive offloading truly persists.

2. On AI’s environmental impact: Of all the concerns we have studied, this is the one we are least comfortable waving away, because the costs are real. Training and running large models consumes meaningful amounts of energy and water, and we do not think it is honest to pretend otherwise. What gives us cautious optimism is the shape of the problem. The vast majority of AI’s energy footprint comes from training these models in the first place, a cost that is paid once by the labs and then spread across billions of uses, which means that a single student asking a question consumes a very small fraction of the total. We also believe efficiency is moving in the right direction, both because smaller and more specialized models are increasingly capable of doing what once required enormous ones and because the shift toward locally-hosted models tends to favor leaner systems that can run on ordinary hardware. None of this makes the impact zero, and we want to be clear that it does not. But we think the responsible path for schools is not to treat AI as uniquely wasteful and ban it, while leaving every other energy-intensive technology in the building untouched, but rather to adopt it deliberately, with an honest accounting of its costs and a preference for the most efficient tools that meet a student’s actual needs. This is the standard we hold ourselves to as well.

3. On student data privacy: While at this point effectively all student messages are being sent to a cloud provider (such as OpenAI, Anthropic, or Google), the world is moving towards locally-hosted models. This means that the LLMs that receive inputted messages run on the messenger’s own hardware, such as their computer. There has been promising research done in this space, and Apple themselves have taken the position that AI will run on user-owned hardware in the near future. For education, this means that while we currently have to trust cloud providers when they say they are not misusing student data, we will likely soon live in a world where this is not a concern at all. Furthermore, even current cloud providers are held to extremely rigorous data privacy standards, and there are massive penalties for any corporations found liable for misusing data, especially in the case of students and minors.

4. On model hallucination: Model hallucination has and will continue to be one of the most challenging issues that AI will face in its adoption. The encouraging news is that the risk is both shrinking and, more importantly, controllable. Newer frontier models hallucinate considerably less than their early predecessors, and peer-reviewed research has shown that the most effective safeguard is, somewhat counterintuitively, not waiting for a perfect model but changing how the model itself is used. When an AI system is grounded in a trusted, curated body of source material, such as a syllabus or assignment, (rather than answering from memory) and is permitted to admit when it doesn’t know something, hallucination rates drop dramatically, in some studies to near zero. We have taken this exact philosophy with our chatbot Sage which exists on Deskpad’s platform Learn. By anchoring responses to vetted educational material and designing for transparency over confident guessing, we can give schools the benefits of AI while keeping the factual reliability that a classroom demands.

We believe that if all of these concerns are addressed honestly and diligently, AI truly can change the way people learn for the better. If we live in a world where education and information is truly democratized, our populations will be smarter, happier, and more likely to create positive change. In every instance in human history where education has been made a top priority, the population has thrived. We believe this moment is no different. Our population can either massively benefit, or massively suffer, from the adoption of AI. If we want to benefit, we must band together as a population, and adopt it on our own terms. This is the sole reason for Deskpad’s existence, and why we believe it is such an important problem to be solving.

Across the country, students, as they tend to do, began immediately exploiting this new tool in the classroom, and before long, teachers were caught in a frenzy of AI-generated submissions that they could not tell apart from genuine student work. The first solution to this problem made its way to the public sphere in January of 2023, just a month and a half after ChatGPT’s initial launch. Princeton undergraduate student Edward Tian launched the AI-detection tool GPTZero, which claimed to “use ChatGPT to detect itself.” Tian had been deeply immersed in natural language processing (NLP) research at Princeton, and his proposed solution seemed like a logical next step for many educators who had gotten used to the Turnitin “gotcha”-style approach to student integrity. For those unfamiliar, Turnitin’s claim to fame that drove them to an estimated $150M+ annualized revenue and an acquisition price of $1.75B in 2019 was the Similarity Score. John M. Barrie, along with collaborators Christian Storm, Emmanuel Briand, and Melissa Lipscom, founded Turnitin in 1998 as PhD candidates at UC Berkeley. Their philosophy was straightforward. Academic cheating—especially plagiarism—had seen a tremendous uptick with the growing adoption of the Internet. Students now had a seemingly infinite list of sources to pull writing from (and pass it off as their own), and educators everywhere were struggling to deduce what work was authentic and what was copied and pasted in from an external source. While some ignored the problem, many were actively searching for a solution, which is exactly where Turnitin came in. The Similarity Score was an algorithm that was designed to compare a student’s work to existing published work on the internet, and generate an integer percentage between 1 and 100 on how likely the work was plagiarized. The solution felt deeply ahead of its time, and for the decades to come, its adoption became widespread in nearly every academic setting.

GPTZero promised educators something familiar. Just as the Similarity Score gave teachers a percentage for plagiarism, GPTZero and the wave of detection tools that followed offered a percentage for AI-generated content. Run the student’s work through the system, get a number back, make a judgment call. For teachers who had spent years relying on Turnitin as their safety net, AI detection felt like the obvious next chapter.

But there was a fundamental problem with this approach that most people didn’t see at first. Turnitin’s Similarity Score works because plagiarism is a matching problem. You’re comparing one piece of text against a database of existing texts and looking for overlap. Either the words match or they don’t. It’s not perfect, but the underlying logic is fairly sound. AI detection is a completely different kind of problem. You’re not comparing text against a source. You’re trying to determine whether the text itself was produced by a human brain or a language model, and you’re doing this by looking at statistical patterns like perplexity and burstiness. Perplexity measures how predictable each word in a sentence is, and burstiness measures how much variation there is in sentence length and structure. The theory is that AI-generated text tends to be more uniform and predictable, while human writing is messier and more varied.

The issue is that these patterns are not reliable indicators of authorship. A student who writes clearly and concisely can produce text with low perplexity that looks identical to AI output. A student who learned English as a second language might write with patterns that a detector interprets as machine-generated. And a student who happens to organize their essay in a structured, methodical way will trigger the same flags that ChatGPT does. In 2023, professors at Texas A&M University accused an entire class of using AI on their final assignments based on detector results. Some students were threatened with failing grades or holds on their diplomas. Many of them had not used AI at all.

Then there’s the other side of the arms race. Within months of GPTZero’s launch, tools like Quillbot, Grammarly Humanizer, and countless others emerged that could take AI-generated text and “humanize” it, introducing enough randomness to fool detectors. The cost of evading detection dropped to zero almost immediately. The students who were actually cheating found workarounds in minutes, while the students who wrote honestly were the ones getting flagged. The system was punishing the wrong people.

What we’re left with is a landscape where schools have been forced to choose between two broken options. Some have banned AI entirely, blocking ChatGPT on school networks and treating any use of it as academic dishonesty. Others have gone the opposite direction, allowing unrestricted use with no guardrails and no visibility into what students are actually doing. Neither approach works, and both are trying to answer the same question: “Did this student use AI?” I think that’s the wrong question. It treats AI use as a light switch, on or off, when in reality there’s a massive spectrum between a student who pastes “write my essay” into ChatGPT and a student who uses AI to pressure-test their own argument before revising. Both of them “used AI.” Only one of them learned anything. The question education should be asking is not whether students are using AI. It’s whether they’re thinking. And right now, nobody has a good way to answer that.