Federal AI push collides with classroom data on romance, risk, and weaker ties

Washington is pressing schools to adopt artificial intelligence. Students are already there. New national data shows 86% of students used AI last school year, and nearly one in five high-schoolers say they or someone they know has had a romantic relationship with it, according to CDT’s 2025 K-12 AI survey.

In April, the White House issued an executive order to promote AI literacy and bring the technology into classrooms. By July, the Education Department followed with a “Dear Colleague” letter telling districts how to use federal grants to support responsible adoption. The First Lady’s office has spotlighted deepfake harassment as part of the same push. The policy posture is clear: move AI from pilot to practice.

The correlation that matters

The new CDT research doesn’t just list problems. It shows they scale with use. Half of all students say using AI in class makes them feel less connected to teachers; in high-use schools, that rises to 56%. Awareness of school-linked deepfakes stands at 36% overall, but reaches 61% in high-adoption environments, where students are also more likely to hear about non-consensual intimate-imagery deepfakes. Teachers in high-use schools are more likely to report large-scale data breaches (28% versus 18%). That’s not background noise. It looks like a dose–response curve.

Behavior outside class tracks the same pattern. Forty-two percent of students say they or a friend used AI for mental-health support, as a friend or companion, or to escape reality. Nineteen percent report a romantic AI relationship in their orbit, with higher rates where schools deploy more AI. CDT’s Elizabeth Laird puts the mechanism plainly: the more tasks a school assigns to AI, the more likely students are to treat AI as a companion or partner. The environment shapes the behavior.

The device vector problem

Here’s where policy meets practice. Nearly one-third of students who converse with AI for personal reasons do so on school-issued devices or software. That blurs the line between district oversight and private life, especially on laptops that run monitoring tools. Students who own personal devices can keep those conversations off the school network; students who rely on district hardware cannot. It’s an equity question as much as a privacy one. The school laptop becomes the gateway.

Monitoring software was meant to keep students safe, yet it often triggers mistaken alerts and scoops up personal information that isn’t related to classwork. Those same devices also put companion-style chat apps one click away, raising hard questions about how much time teens spend with them—and what that means for real-world relationships.

The training vacuum

Teachers are embracing the tools—many to plan lessons or tailor materials—and 47% report receiving some training on classroom use, with most of those teachers finding it helpful. But that support rarely covers student wellbeing. Only 11% say they were coached on what to do if a student’s AI use looks harmful. Students say they rarely get clear rules about when AI belongs in an assignment or what “responsible use” actually means. Many teachers are uneasy, too. Roughly seven in ten think heavy classroom AI erodes basics like reading and writing, and most worry it squeezes the person-to-person skills schools are supposed to build. The result is predictable: usage rises faster than the ability to steer it.

Federal principles vs. school realities

Federal guidance now lists five principles for “responsible use”: educator-led, ethical, accessible, transparent, and data-protective. But principles in a letter are only as strong as the strings attached to money. That’s why nine civil-society groups are urging the department to embed those standards into grant conditions and research programs authorized by the executive order. “Should” needs to become “must” if procurement is going to change. Districts buy to the spec that funding requires. Everything else is optional.

There’s also a measurement gap. District dashboards track adoption and device uptime; few track the human side. If leaders want to know whether AI is helping, they need to monitor connection to teachers, counseling referrals, exposure to deepfakes, and the share of personal AI use happening on school devices. Absent that, debates reduce to hunches.

States are starting to legislate

Some states are moving first. Ohio has set deadlines: districts must adopt an AI use policy by July 1, 2026, and the state education department plans to release a model policy before year’s end. In Tennessee, lawmakers are considering a measure that would require grades 6–12 teachers to finish a state-approved AI course by 2027, with the training provided at no charge. These moves matter because mandates, not memos, alter vendor roadmaps and district timelines. When a statute says “adopt a policy,” boards vote. When a statute says “train teachers,” schedules change and budgets follow.

Elsewhere, officials are experimenting with softer levers—model policies, voluntary guidance, and suggested classroom norms. Those help, but they rarely survive a tight budget or a crowded agenda. The districts most eager to adopt AI will push ahead; the districts most cautious will wait. The middle will drift. Rules close the gap.

What to watch next

Three levers will decide whether schools course-correct. First, whether the Education Department ties its principles to real funding requirements—publish a district AI policy, turn off companionship features in school accounts, verify age-appropriate defaults, and grant local audit rights. If those conditions appear, schools will implement them fast. If not, adoption will keep outrunning governance.

Second, whether statehouses copy Ohio’s mandate model or settle for training-only approaches that leave procurement unchanged. Policies create selection pressure; professional development alone changes habits slowly.

Third, whether districts start measuring what matters. Build simple, public dashboards that track student–teacher connection, time on task, counseling demand, and incidents tied to deepfakes or chatbot dependence. If those indicators worsen as AI use intensifies, slow deployment until protections catch up. That’s not anti-technology. It’s basic risk management.

The trade-offs are now visible. The benefits—faster lesson prep, individualized practice, always-on help—are real. So are the costs—thinner relationships, more deepfakes in school communities, and rising reliance on systems that imitate closeness. The federal government is stepping on the gas. Students are asking for brakes. Schools shouldn’t wait for a headline to make the case.

Why this matters:

• Districts are scaling intimate AI systems without matching safeguards, and the documented harms rise as adoption intensifies.
• Attaching enforceable conditions to federal and state funds—policy, training, device rules, and measurement—will decide whether schools buy “principled” AI or simply more AI.

Students Use AI; Schools Delay—Big Tech Writes the Rules

Students embrace AI faster than schools can write rules. While 85% use AI for coursework, institutions stall on policy—and tech giants step in with billions in training programs to fill the vacuum. The question: who gets to define learning standards?

Implicator.aiMarcus Schuler

Most US Teens Now Confide in AI Instead of Humans

72% of US teens use AI companions for emotional support and serious conversations. 31% find AI chats as satisfying as human relationships.

Implicator.aiMarcus Schuler

Most US Teens Now Confide in AI Instead of Humans

Most parents worry about teens pulling away during adolescence. They don’t expect kids forming intimate bonds with AI. 72% of US teens now use AI companions for emotional support, flirting, and serious conversations they’d normally have with humans.

Implicator.aiMaria Garcia