# Handed the Pen, Students Chose Restrictions: The First Student AI Policy > Source: > Published: 2026-08-10 01:00:30+00:00 ## Handed the Pen, Students Chose Restrictions: The First Student AI Policy Campbell McHenry, a high school junior from Mississippi, had a word for what happens to a school with no AI policy. Unhinged. “Everyone's gonna stop learning and start relying on AI,” he told NPR during a weekend in Boston in July. “Instead of using it as a tool, they're gonna use it as a brain.” He was sitting in a 5,500 square foot replica of the United States Senate chamber, on the campus of the University of Massachusetts Boston, built by the Edward M. Kennedy Institute for the United States Senate. It is the only full-scale copy of the chamber in the world. The desks are accurate. The rostrum is accurate. The lighting is accurate. What was not accurate, over the weekend of 17 to 19 July 2026, was the age of the people in the seats, and the fact that they were getting something done. Ninety-eight high school students, mostly juniors and seniors, one or two from each state, spent thirty-eight hours arriving by 42 inbound flights, by Amtrak, by bus and by car, in order to draft and pass a piece of legislation about artificial intelligence in K-12 schools. On the Sunday morning they voted. The STUDENTS FIRST Act passed 82 to 16. It is not a law. The students are not senators. The whole thing was a simulation, an expanded version of the Kennedy Institute's Senate Immersion Module, the role-play programme that has run more than 160,000 students from 21 states through the replica chamber, and which in autumn 2024 added technology reform to its slate of debate topics. This particular weekend was branded America's Youth AI Festival and organised with Day of AI, the nonprofit spun out of MIT's RAISE initiative, and AASA, The School Superintendents Association. And yet the text those teenagers produced is now going somewhere real. AASA has said it will send the bill to its membership: more than 10,000 school system leaders, a great many of whom are, at this moment, writing AI policies of their own with almost nothing to copy from. Which raises a question that ought to be more uncomfortable than it has been. Why is the most specific, most internally coherent piece of K-12 AI legislation in circulation this summer a document written by children? ## What Ninety-Eight Teenagers Actually Wrote Down The bill is real enough to read, and reading it is more instructive than reading the coverage of it. AASA has published the full nine-page text as S.2026, styled after the 119th Congress. The acronym is tortured in the authentic congressional manner: Safe Technology Use, Data Privacy, and Ethical Norms for Teaching, Fostering Integrity and Responsible Systems in Teaching. It runs to five sections, addressed in turn to everyone, to students, to parents and guardians, to teachers and to administrators. Each section after the first was worked up by a subcommittee with its own strained acronym, then amended and adopted by the full body. The student section is the longest, at fifteen provisions. What is striking, on the page, is how much of the bill points away from students. Section one opens not with a rule about teenagers but with a rule about vendors: any organisation selling generative AI services to public schools “must address all known biases within their generative AI, and must inform their clients of the biases that their products represent,” and the school must then pass that information to parents. Section four tells teachers that “just as students are expected to be transparent in their AI usage, teachers should be held to the same standard.” Section five forbids administrators from using AI to hire staff, forbids them from entering student or teacher personal information into AI tools, and requires them to “consult with students, staff, and families before drafting AI rules.” There is a running preoccupation with human contact. Educators, section one says, should omit AI “in situations where human connection is irreplaceable in the learning environment.” Administrators should require that school counsellors do not use AI tools, “particularly in mental health contexts.” Nothing in the act, it adds, shall penalise a teacher for declining to use AI at all. This is not the document you would predict if you assumed teenagers writing their own AI rules would write themselves a permission slip. It reads instead like something drafted by people who have concluded that the technology's main risk to them is not that they will be caught using it, but that the institution around them will be hollowed out by it. ## A Theory of Assessment Sharper Than Most Districts Have The most consequential thing in the bill is also the easiest to skim past. Provision nine of the student section: “AI use is strictly prohibited on all graded assessments.” Then, immediately: “AI use is permitted for studying.” That is a theory of assessment, and a defensible one. It says AI is a legitimate part of learning and an illegitimate part of measurement. Use it to get better; do not use it at the moment we are checking how good you got. The students arrived at this in a weekend. Australian higher education took rather longer to arrive somewhere similar. The two-lane approach set out by Adam Bridgeman and Danny Liu at the University of Sydney in July 2024, and since widely discussed across the sector, splits assessment into a secure lane, conducted under controlled conditions to verify learning, and an open lane in which AI use is scaffolded and taught. Australia's tertiary regulator followed in September 2025 with guidance on enacting assessment reform, built on the premise that detection cannot guarantee integrity and that structural redesign is the only durable answer. Most American district policies have not made this distinction cleanly. They tend to produce a general posture, permissive or restrictive, and leave the assessment question to individual teachers. The student bill instead builds a scaffold: teachers must publish a written AI policy each semester, “explicitly stating when AI is permitted, restricted, or banned for specific assignments,” with “category based rules (essay, worksheet, test),” and those rules must be consistent across sections of the same course. The students, in other words, identified the thing that drives them mad, which is not that AI rules exist but that they differ from classroom to classroom and are frequently unwritten, and legislated against it. They also drew a line most adults have avoided. Provision four: students “should never use AI for the act of writing,” though editing, brainstorming and studying with it are permitted after eighth grade, and only after non-AI editing has been “taught and assessed thoroughly.” Whether that is the right line is arguable. That it is a line, stated in one sentence, with an age attached and a prerequisite skill named, puts it ahead of a great deal of adult guidance. ## The Clause Nobody Asked Them For Then there is the carbon. Provision two of the student section requires states to mandate AI literacy instruction from the moment students start using devices in the classroom, and specifies the syllabus: misinformation, plagiarism, bias in AI systems, privacy concerns, appropriate academic use, how AI functions, and “environmental impacts.” Nothing in the American federal framework centres that last item, largely because the American federal framework barely exists. Executive Order 14277, signed on 23 April 2025 and titled Advancing Artificial Intelligence Education for American Youth, established a White House Task Force on AI Education, directed the Education Secretary to issue guidance on how existing grant funds might be spent, told the National Science Foundation to prioritise research, and launched a Presidential AI Challenge. Its subject matter is capability and workforce readiness. It does not address academic integrity. It does not address assessment. It does not mention the energy and water cost of the systems it wants children to become proficient with. The students put it in the curriculum unprompted, alongside plagiarism, as a thing you are supposed to know about the tool before you are handed it. It is a small provision and it tells you something about the frame these teenagers are working in. They are not treating AI as a neutral productivity layer that has arrived to help them. They are treating it as an industrial product with costs, made by companies with interests, and they want the cost sheet taught in class. The literacy mandate itself was the weekend's most contested item, and the debate was not sentimental. It was about money. Om Lala of California objected that requiring schools to teach AI literacy would be a heavy lift because “teachers already are pretty under-resourced.” Tristan Tjetjep of Tennessee worried about how “underprivileged schools” would pay for it. The students had walked directly into the structural funding inequity of American public education, and they responded the way legislatures respond, by writing a funding requirement into the bill: proportional federal funding with a baseline minimum, for enterprise AI, mandatory staff training and a range of student literacy options. They invented an unfunded mandate and then, within the same weekend, tried to fund it. ## Viva Voce and the Burden of Proof The most legally fraught thing in the bill sits in section four, provision six, under a Latin heading. “VIVA VOCE (Oral Defense): Teachers reserve the right to require an oral defense of any paper or project where AI use is suspected or reported, to ensure the student possesses the knowledge displayed in the work.” Take the pragmatic case first, because it is strong. AI detectors do not work well enough to hang a disciplinary finding on. The most cited evidence is a 2023 study in the journal Patterns by Weixin Liang and colleagues at Stanford, which ran seven widely used GPT detectors over essays written by non-native English speakers and found that more than half of the TOEFL essays were misclassified as AI-generated, while the same detectors were near-perfect on writing by US eighth graders. The mechanism is banal and damning: detectors flag low linguistic variability, and non-native writers, along with anyone writing plainly, produce exactly that. The students knew this. Their bill says a suspected student is entitled to review by two school officials as well as detection software, and that teachers “must personally investigate the issue before reporting the student to admin.” They wrote in a right to appeal any finding of wrongful AI use, and a right of appeal against any disciplinary action. That is more due process than a lot of real districts currently offer, and it was written by people who expect to be on the receiving end of it. Microsoft's 2025 education survey found the single largest concern among American higher education students was not misinformation or job loss but “potential to be accused of plagiarism or cheating,” named by 32 per cent. But the viva provision still has a problem, and it is the direction the burden runs. An oral defence triggered by suspicion asks a student to prove authorship rather than asking the school to prove misconduct. Perform well and you are cleared; perform badly and the suspicion is confirmed. The evidence on oral assessment is not reassuring here. A systematic review of the structured viva in health professions education found anxiety to be the most prominent reported limitation, along with weak standardisation and equity problems for students speaking English as an additional language. Layer that on a student with an anxiety disorder, a speech difference, a language barrier, or an individualised education programme, and a mechanism designed to protect the wrongly accused becomes a second way to fail them. To their credit, the students left an escape hatch of sorts: their own bill requires schools to allow AI-based learning supports in IEP and 504 accommodation plans, and advises students to preserve document history on major assignments as a defence against false positives. That last piece of advice is the most telling line in the entire document. It is a generation coaching itself on how to prove it did its own homework. ## What the Adults Produced Instead Set the student bill beside the adult record and the asymmetry is the story. There is no national framework for AI in American schools. The April 2025 executive order imposes no obligation on any district. At state level, the picture is guidance-heavy and law-light: by late October 2025, 34 states plus Puerto Rico had issued some form of official K-12 AI guidance, according to the tracker maintained by AI for Education. Guidance is not requirement. An analysis published by ExcelinEd in May 2026 counted just seven states with enacted laws, another twenty or so with bills in progress, and around ten with nothing beyond guidance or a task force. Its summary of the gap was that districts are being left to vet tools without any state-provided evaluation framework. Congress, meanwhile, spent 2025 trying to make the state patchwork illegal. A ten-year moratorium on state AI regulation was inserted into the budget reconciliation bill, and on 1 July 2025 the Senate voted 99 to 1 to strip it out. The single vote in favour of preemption belonged to Thom Tillis. Federal AI policy for the year was, in substance, a near-unanimous decision not to have one, and to let the states carry on. So when 10,000 superintendents open an email containing a bill written by teenagers, and from August sit through webinars and school visits built around it, they will be doing so because nothing else has arrived. Jeffrey Riley, who runs Day of AI and was previously Massachusetts commissioner of elementary and secondary education, put the logic plainly to NPR: “The adults themselves haven't been able to come to consensus on this, right? And so now it falls to the kids. Let them lead the way.” That is a generous reading. A less generous one is that adult institutions have discovered that student voice is cheap, fast and reputationally safe in a way that binding policy is not. Nobody gets sued over a simulation. ## The Ghost of Social Media in the Chamber The students had a theory about why the adults were slow, and they said it out loud all weekend. “We missed the mark with social media and phones,” said Hunter Wurzel, a senior representing Ohio. “That's been a huge talking point already, is how we failed to implement those things into education and teach the youth how to use them.” It is worth taking that claim apart rather than nodding at it, because the record is more specific than the slogan. The Surgeon General issued a formal advisory on social media and youth mental health in May 2023, noting that up to 95 per cent of teenagers aged 13 to 17 used a social platform and more than a third said they used one “almost constantly.” In June 2024 the Surgeon General went further and called for a tobacco-style warning label. That step requires an act of Congress. Congress did not act. The Kids Online Safety Act passed the Senate 91 to 3 on 30 July 2024, then died without a House vote. Reintroduced in the 119th Congress, it was eventually folded into a House package, the KIDS Act, which passed 267 to 117 on 29 June 2026, with KOSA's central duty of care stripped out. Senate co-author Richard Blumenthal promptly declared the House version dead in his chamber. Two years and one week after a 91 to 3 Senate vote, there is still no statute. Phone bans tell a similar story at a different pace. As of the end of April 2025, only eleven states had statewide restrictions of any kind, four of them full bell-to-bell bans, with Florida first in 2023. Fourteen months later the position had transformed: by June 2026, 22 states and Washington D.C. had adopted full bell-to-bell bans, and a further 19 had enacted more flexible legislation leaving districts to set their own policies, putting roughly 41 states under some form of restriction. Common Sense Media's own framing of the arms race is the sharpest version: it notes that Pew found roughly 73 per cent of online American teenagers were using social networking sites by 2009, about three years after Facebook opened to the public. The policy response to a technology that saturated adolescence by 2009 became widespread state law around 2025. This is the founding political fact for the cohort in that chamber. Every one of those 98 students was born into a live, uncontrolled experiment, watched adults spend fifteen years arguing about whether it was harmful, and reached the age of legislative role-play just in time to be handed the next one. “We missed the mark” is not adolescent grandstanding. It is a reasonably precise summary of the legislative record, delivered by the test subjects. ## The Evidence Behind the Fear They Named McHenry's line about using AI as a brain rather than a tool is a folk statement of a hypothesis researchers are actively testing, with results that are suggestive rather than settled. The most publicised is the MIT Media Lab study led by Nataliya Kosmyna, posted to arXiv in June 2025 under the title “Your Brain on ChatGPT,” which had 54 participants write SAT-style essays across three sessions while wearing 32-electrode EEG headsets, with 18 returning for a fourth crossover session. Brain connectivity scaled down with the amount of external help: unaided writers showed the strongest networks, search engine users intermediate, language model users the weakest. Participants who had used the model struggled to quote their own essays back and reported weaker ownership of them. The authors called the effect cognitive debt. It is a preprint with a small sample and a narrow task, and it has been over-read constantly since publication. It is not proof that AI makes teenagers stupid. What it does establish is that the mechanism the students are worried about is at least physically detectable. The correlational picture points the same way without proving causation. Michael Gerlich's 2025 study in Societies, surveying 666 participants, found a significant negative association between frequent AI use and critical thinking scores, mediated by cognitive offloading, with the strongest dependence among 17 to 25 year olds. Common Sense Media's 2026 census of 1,204 American children aged 9 to 17, fielded with SSRS on a probability panel, found that 20 per cent of AI users, and 42 per cent of daily users, said going a month without AI would be very or somewhat hard. It also found that children who struggle academically lean hardest on it: 56 per cent of those who have a hard time staying focused on assignments use AI for schoolwork at least weekly, against 45 per cent of those who do not. That last finding is the uncomfortable one, because it cuts against the student bill's own logic. A restriction-first regime falls heaviest on the students already struggling, who are using the tool most, and who are least likely to have the document history, the articulacy under pressure, or the parental advocacy to survive an oral defence. ## Ninety-Eight Is Not a Sample There is a version of this article that treats the Students First Act as the authentic voice of American youth. That version would be wrong, and the numbers say so. These students were, in NPR's phrasing, “handpicked by their school leaders.” They are the young people a superintendent thinks of when asked to nominate someone for a civics simulation in Boston: engaged, articulate, available in July, backed by families who can get them to an airport. Selecting for civic enthusiasm selects for institutionalism. A cohort assembled that way is likely to believe rules can work, because believing that is roughly the entry requirement. Set them beside the general population and the gap opens. The Common Sense census found 86 per cent of 9 to 17 year olds use AI in some form, rising to 92 per cent of 16 and 17 year olds, with 24 per cent using it daily. Among users, 85 per cent use it for schoolwork, 48 per cent at least weekly and 21 per cent every day. Asked in an open-ended question what people their age know about AI that adults do not, the single largest theme, at 21 per cent, concerned schoolwork, including how to cheat with it and, in the researchers' summary, “how kids can get away with using AI for schoolwork even when they are not supposed to.” One respondent's verbatim answer: “You can exploit AI to do all your assignments from school, and use multiple AIs to make it look handmade.” That is the constituency the Students First Act would govern, and it is not the constituency that wrote it. There is also the performance problem. A room of teenagers, in a Senate chamber, watched by superintendents, a former state education commissioner and a national broadcaster, is a room with a strong incentive to produce responsibility. The bill's restrictionism may be partly sincere and partly a read of the audience. Both can be true. Adult legislators are not usually granted the assumption that their public positions are unaffected by who is watching, and the students should not be either. What complicates the cynical reading is the texture of the dissent. Jadelynn Petitjean of Alabama called the bill an “unstable foundation” and voted no. Ethan Liu of West Virginia catalogued the contradictions, including amendments that restricted student access while handing teachers broad discretion, and told the chamber he had “a whole different speech planned out against this bill, actually. It was really good.” He changed his vote only after organisers let the students patch some of the inconsistencies. Zora Nobel of Maine voted for it while describing it on the record as containing “repetitive provisions, prescriptive language, and at times, contradictions.” That is not a room performing consensus. That is a room bargaining. ## Why Rules You Wrote Yourself Might Actually Hold The strongest argument for the exercise has nothing to do with the quality of the text. Tom Tyler's work, running from Why People Obey the Law onward, is the foundation of procedural justice theory, and its central finding is that compliance tracks perceived legitimacy far more reliably than it tracks fear of punishment. People follow rules they consider fairly made by authorities they consider entitled to make them. Tyler's corollary is the one that should worry school administrators: regulating behaviour primarily through threat can actively erode the internal commitment that produces voluntary compliance. Research on adolescents specifically finds that transparency of rationale and genuine consideration of young people's views have consistent effects on trust in authority, in schools and in government alike. Applied here, the argument is not that teenagers wrote better AI rules than superintendents would have. It is that a rule a student helped write is a rule a student is more likely to keep, and that AI compliance is almost entirely a voluntary-compliance problem, because the enforcement technology does not work. The research literature has begun to catch up. A peer-reviewed paper by Kaoru Seki, Manisha Vijay and Yasmine Kotturi, titled “Participatory, not Punitive,” presented at the CHI conference on human factors in computing systems in Barcelona in April 2026, ran a three-part workshop series in a graduate design course at a minority-serving American university, in which two student leaders facilitated discussion with no faculty in the room. The participants co-authored ten policy recommendations. The authors argue that engaging students in AI governance carries value beyond the resulting policies, and describe their model as calling students in rather than calling them out. It is a small study, eight participants, at university rather than K-12, and should not be asked to carry more than it can. But one of its specific findings maps onto the Boston bill with uncanny precision. The student-led process surfaced a concern absent from top-down governance: the double standard of requiring students to disclose or abstain from AI while faculty face no equivalent expectation. Section four, provision four of the Students First Act, written by teenagers who had not read that paper, says teachers must be held to the same transparency standard as students. Two independent participatory processes, different countries of the academy, different age groups, produced the same demand. That is the sort of convergence that suggests the process is surfacing something real rather than flattering its participants. ## When the Text Meets a School Board None of this survives contact with a school board meeting unaltered, and it is worth being specific about where it breaks. Start with enforceability. “AI use is strictly prohibited on all graded assessments” is a rule about a device, not a rule about a mind, and the bill's own enforcement mechanism, mandating “one standardized AI detector” per district plus lockdown browsers, leans on precisely the technology the students elsewhere acknowledge is unreliable. There is an unresolved contradiction here: the bill requires teachers to use a detector on written graded assessments while simultaneously requiring that no student be judged on a detector alone. Both things can be operationalised, but only by a district with the staffing to run a two-official review on every flag. That district exists. So does the one with a single overstretched assistant principal. Then the funding. The bill's answer to inequity is proportional federal funding with a baseline minimum, which is a sentence a student senate can write and a school board cannot. Absent that money, an AI literacy mandate starting at the moment of device issue becomes another compliance burden distributed unequally, heaviest on the districts with the least capacity, which is the exact pattern the students were trying to prevent. Then the parents. Section three is the weakest part of the bill, and the students knew it. “We cannot legislate parenthood,” Isabella Stringer of New Hampshire told the chamber, and the section duly softens into advice. Except in one place, provision seven, where a parent using AI on a child's graded work without teacher permission exposes the child to suspension or expulsion. That will not survive a single angry school board meeting, and it should not. The record-keeping is imperfect too, in a way that is almost endearing. The PDF AASA published is filed under a name declaring the bill passed 84 to 14. The page hosting it, NPR's account and AASA's own national announcement in August all say 82 to 16. Somewhere between the chamber and the content management system, four votes moved. Real legislatures have the same problem and solve it with a clerk. ## The Line They Drew The temptation is to end warmly. Ninety-eight teenagers did in a weekend what Congress has not done in three years, and they did it with more precision about assessment, more attention to due process and more interest in the environmental cost of the technology than the sitting federal framework contains. The temptation in the other direction is equally available. It was a simulation, in a replica, with a non-binding output, at an event underwritten in part by an AI literacy nonprofit, involving students selected for their enthusiasm, producing a document with acknowledged internal contradictions that will be put in front of superintendents who are under no obligation to adopt a word of it. Laura Lundy, writing in the journal Childhood in 2018, made the case that the children's rights field has been too quick with the word tokenism. Since Roger Hart placed tokenism on the bottom rungs of his 1992 ladder of children's participation, alongside manipulation and decoration, the field has often concluded that imperfect participation is worse than none. Lundy's argument, drawn from what participating children actually report, is that this reasoning is used to justify doing nothing, and that young people who have taken part in flawed processes frequently do not experience them as empty. The useful question is not whether the Boston weekend was real politics. It is what the students chose when handed the pen, and that answer is not ambiguous. Offered an opportunity to write themselves permission, a cohort that uses this technology more than any group in history wrote itself restrictions, then wrote in appeal rights, then turned round and imposed transparency obligations on the vendors, the teachers and the administrators too. They banned AI on the tests and allowed it in the studying. They asked to be taught what the machines cost the planet. They asked, in the last provision of the bill's last section, to be consulted before adults write rules about them. Bena Mejia Romero of Connecticut put it to the chamber before the final vote: “This bill is about who gets a seat at the table. Today, students have that seat.” It was, for that weekend, a seat in a room where the desks were accurate and the votes were not counted by anyone who had to act on them. On 3 August 2026 the organisers announced what happens next, and it is more than an email. AASA will put the Act in front of its full membership, and Day of AI, the Kennedy Institute and MIT's RAISE initiative will follow it into schools with webinars, workshops, public discussions and in-person visits. Sarah Yezzi, the Kennedy Institute's chief educator officer, summarised the students' position in that announcement: “Students are not asking schools to ignore AI, nor are they asking schools to adopt AI without limits. They want clear, fair rules that protect privacy, preserve the ability to think, and ensure this transformative technology supports and enhances human relationships.” A webinar is not a statute and a school visit is not a school board vote. Nothing in the programme obliges a single district to adopt a single provision, and the bill remains what it was on the Sunday morning it passed: a text with no force behind it but its own reasoning. What the announcement does dispose of is the easy dismissal, which was that the thing would land in ten thousand inboxes and stay there. It is going to be argued about in rooms where the votes are counted by people who have to act on them. The question is no longer whether anyone reads it. It is whether any of it ever becomes binding on anyone. ## Sources and References - Cory Turner, “Adults have struggled to set rules for AI in school. These teens figured it out,” NPR, 30 July 2026. [https://www.npr.org/2026/07/30/nx-s1-5853571/students-set-ai-policy](https://www.npr.org/2026/07/30/nx-s1-5853571/students-set-ai-policy) - AASA, The School Superintendents Association, “The Students First Act: A Student-Led AI Policy,” 30 July 2026. [https://www.aasa.org/resources/resource/students-first-act](https://www.aasa.org/resources/resource/students-first-act) - 2026 Student Senate, “S.2026, The STUDENTS FIRST Act,” full bill text published by AASA, 19 July 2026. [https://www.aasa.org/docs/default-source/resources/reports/17_s-2026-the-students-first-act-final-bill-passed-84-14-july-19-2026.pdf](https://www.aasa.org/docs/default-source/resources/reports/17_s-2026-the-students-first-act-final-bill-passed-84-14-july-19-2026.pdf) - AASA, The School Superintendents Association, “Students from All 50 States Produce National Framework for AI in America's Schools,” 3 August 2026. [https://www.aasa.org/news-media/news/2026/08/03/students-produce-national-framework-for-ai-in-america's-schools](https://www.aasa.org/news-media/news/2026/08/03/students-produce-national-framework-for-ai-in-america's-schools) - Edward M. Kennedy Institute for the United States Senate, “Edward M. Kennedy Institute introduces a new AI policy program to support civic and digital literacy,” 2024. [https://emkinstitute.org/about-the-institute/edward-m-kennedy-institute-introduces-a-new-ai-policy-program-to-support-civic-and-digital-literacy/](https://emkinstitute.org/about-the-institute/edward-m-kennedy-institute-introduces-a-new-ai-policy-program-to-support-civic-and-digital-literacy/) - Kaoru Seki, Manisha Vijay and Yasmine Kotturi, “Participatory, not Punitive: Student-Driven AI Policy Recommendations in a Design Classroom,” Proceedings of the 2026 CHI Conference on Human Factors in Computing Systems (CHI '26), Barcelona, 13-17 April 2026, DOI 10.1145/3772318.3790691; preprint arXiv:2604.10851. [https://doi.org/10.1145/3772318.3790691](https://doi.org/10.1145/3772318.3790691) - Harvard Kennedy School, “Many states are banning cellphones in schools. Is it working?,” featuring Paul Peterson, 9 June 2026. [https://www.hks.harvard.edu/faculty-research/policy-topics/education-training-labor/many-states-are-banning-cellphones-schools](https://www.hks.harvard.edu/faculty-research/policy-topics/education-training-labor/many-states-are-banning-cellphones-schools) - Microsoft Education, “2025 AI in Education Report: Extended Survey Data,” August 2025. [https://cdn-dynmedia-1.microsoft.com/is/content/microsoftcorp/microsoft/bade/documents/products-and-services/en-us/education/Survey-Data-2025-Microsoft-AI-in-Education-Report.pdf](https://cdn-dynmedia-1.microsoft.com/is/content/microsoftcorp/microsoft/bade/documents/products-and-services/en-us/education/Survey-Data-2025-Microsoft-AI-in-Education-Report.pdf) - Eleanor O'Neil, Kristen Soltis Anderson, Baxter White, Laura Zimmermann, Supreet Mann and Michael B. Robb, “The Common Sense Media Census: AI Use by Tweens and Teens, 2026,” Common Sense Media, 2026. [https://www.commonsensemedia.org/sites/default/files/research/report/2026-ai-use-by-tweens-and-teens-1.pdf](https://www.commonsensemedia.org/sites/default/files/research/report/2026-ai-use-by-tweens-and-teens-1.pdf) - Holland & Knight, “Executive Order: Advancing Artificial Intelligence Education for American Youth,” analysis of Executive Order 14277 of 23 April 2025, April 2025. [https://www.hklaw.com/en/insights/publications/2025/04/executive-order-advancing-ai-education-for-american-youth](https://www.hklaw.com/en/insights/publications/2025/04/executive-order-advancing-ai-education-for-american-youth) - The White House, “AI Education,” ai.gov initiatives page, accessed 2 August 2026. [https://www.ai.gov/initiatives/education](https://www.ai.gov/initiatives/education) - Nataliya Kosmyna, Eugene Hauptmann, Ye Tong Yuan, Jessica Situ, Xian-Hao Liao, Ashly Vivian Beresnitzky, Iris Braunstein and Pattie Maes, “Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task,” arXiv:2506.08872, 10 June 2025. [https://arxiv.org/abs/2506.08872](https://arxiv.org/abs/2506.08872) - Weixin Liang, Mert Yuksekgonul, Yining Mao, Eric Wu and James Zou, “GPT detectors are biased against non-native English writers,” Patterns, 10 July 2023. [https://pubmed.ncbi.nlm.nih.gov/37521038/](https://pubmed.ncbi.nlm.nih.gov/37521038/) - Michael Gerlich, “AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking,” Societies, 3 January 2025. [https://www.mdpi.com/2075-4698/15/1/6](https://www.mdpi.com/2075-4698/15/1/6) - Abdelhamid Ibrahim Hassan Abuzied and Wisal Omer Mohamed Nabag, “Structured viva validity, reliability, and acceptability as an assessment tool in health professions education: a systematic review and meta-analysis,” BMC Medical Education, 2023. [https://pmc.ncbi.nlm.nih.gov/articles/PMC10369684/](https://pmc.ncbi.nlm.nih.gov/articles/PMC10369684/) - Adam Bridgeman and Danny Liu, “Frequently asked questions about the two-lane approach to assessment in the age of AI,” Teaching@Sydney, University of Sydney, 2 July 2024. [https://educational-innovation.sydney.edu.au/teaching@sydney/frequently-asked-questions-about-the-two-lane-approach-to-assessment-in-the-age-of-ai/](https://educational-innovation.sydney.edu.au/teaching@sydney/frequently-asked-questions-about-the-two-lane-approach-to-assessment-in-the-age-of-ai/) - AI for Education, “State AI Guidance for Education,” updated 28 October 2025. [https://www.aiforeducation.io/ai-resources/state-ai-guidance](https://www.aiforeducation.io/ai-resources/state-ai-guidance) - ExcelinEd, “State K-12 AI Policy in 2026: Milestones, Momentum and Missing Links,” 26 May 2026. [https://excelined.org/2026/05/26/state-k-12-ai-policy-in-2026-milestones/](https://excelined.org/2026/05/26/state-k-12-ai-policy-in-2026-milestones/) - Goodwin Procter, “Federal AI Moratorium Dies on the Vine as Senate Passes the One Big Beautiful Bill Act,” July 2025. [https://www.goodwinlaw.com/en/insights/publications/2025/07/alerts-practices-aiml-federal-ai-moratorium-dies-on-the-vine](https://www.goodwinlaw.com/en/insights/publications/2025/07/alerts-practices-aiml-federal-ai-moratorium-dies-on-the-vine) - Office of the U.S. Surgeon General, “Social Media and Youth Mental Health: The U.S. Surgeon General's Advisory,” U.S. Department of Health and Human Services, 23 May 2023. [https://www.hhs.gov/sites/default/files/sg-youth-mental-health-social-media-advisory.pdf](https://www.hhs.gov/sites/default/files/sg-youth-mental-health-social-media-advisory.pdf) - Crowell & Moring, “House Passes KIDS Act H.R. 7757: What Online Platforms Must Know Before Senate Showdown,” July 2026. [https://www.crowell.com/en/insights/client-alerts/house-advances-bipartisan-kids-online-safety-bill-but-senate-showdown-looms](https://www.crowell.com/en/insights/client-alerts/house-advances-bipartisan-kids-online-safety-bill-but-senate-showdown-looms) - KFF, “A Look at State Efforts to Ban Cellphones in Schools and Implications for Youth Mental Health,” 2025. [https://www.kff.org/mental-health/issue-brief/a-look-at-state-efforts-to-ban-cellphones-in-schools-and-implications-for-youth-mental-health/](https://www.kff.org/mental-health/issue-brief/a-look-at-state-efforts-to-ban-cellphones-in-schools-and-implications-for-youth-mental-health/) - Roger A. Hart, “Children's Participation: From Tokenism to Citizenship,” UNICEF International Child Development Centre, 1992, summarised by Organizing Engagement. [https://organizingengagement.org/models/ladder-of-childrens-participation/](https://organizingengagement.org/models/ladder-of-childrens-participation/) - Laura Lundy, “In defence of tokenism? Implementing children's right to participate in collective decision-making,” Childhood, volume 25 issue 3, 2018. [https://pureadmin.qub.ac.uk/ws/portalfiles/portal/148759327/Tokenismfin.pdf](https://pureadmin.qub.ac.uk/ws/portalfiles/portal/148759327/Tokenismfin.pdf) - Tom R. Tyler, “Why People Obey the Law,” Princeton University Press, 2006 edition. [https://press.princeton.edu/books/paperback/9780691126739/why-people-obey-the-law](https://press.princeton.edu/books/paperback/9780691126739/why-people-obey-the-law) **Tim Green** *UK-based Systems Theorist & Independent Technology Writer* Tim explores the intersections of artificial intelligence, decentralised cognition, and posthuman ethics. His work, published at [smarterarticles.co.uk](https://smarterarticles.co.uk), challenges dominant narratives of technological progress while proposing interdisciplinary frameworks for collective intelligence and digital stewardship. His writing has been featured on Ground News and shared by independent researchers across both academic and technological communities. **ORCID:** [0009-0002-0156-9795](https://orcid.org/0009-0002-0156-9795) **Email:** [tim@smarterarticles.co.uk](mailto:tim@smarterarticles.co.uk) Listen to the free weekly [SmarterArticles Podcast](https://www.smarterarticles.fm)