Who Decides Which AI Tools Are Safe for Your Students?

Every parent, administrator, and policymaker should ask one simple question right now. Who decides which AI tools are safe for your students? Ask around, and you will mostly get shrugs. AI adoption is moving fast, but our governing systems are moving slowly. That gap is one of the biggest untold stories in American education today.

Part of the problem is basic. No standard definition exists for a “safe” AI tool in classrooms. At a minimum, safe AI tools protect student data privacy. They also produce accurate information and clearly support student learning. Equity matters too, so the tool must limit algorithmic bias. It should also work for all students, including those with disabilities. Finally, teachers need real control over how the tool gets used.

Keep that full definition in mind, because it matters for what comes next.

The Vacuum That Created the AI Tools Problem

ChatGPT launched in late 2022, and schools hit an immediate crisis. Teachers and students were already using it. But nobody had a policy. As of 2023, not a single state had formal AI guidance for K-12 schools.

California moved early and published guidance in late 2023. It encouraged educators to use AI, yet it warned that no tool replaces a caring teacher. Oregon followed with a guide focused on equity and data privacy. Notably, Oregon told educators to stay skeptical of AI detection tools. Those tools have proven unreliable.

Both documents were purely advisory. There was no enforcement, no tool vetting, and no binding rule for districts.

By April 2025, at least 28 states had published some form of AI guidance. Today that number has grown past 30. That is real progress. However, guidance documents do not vet AI tools. They cannot tell a principal at a rural Title I school what to do. She still needs to know if a specific tool is safe and effective. That decision falls to individual districts. Most lack the staff, time, or expertise to make it rigorously.

So a governance vacuum formed. And something always rushes into a vacuum. Here, private consortiums, nonprofit certification programs, and viral teacher adoption filled the space. You can see the same pattern in our look at state technology policies for the new school year.

Who Vets AI Tools Today

Four organizations now shape which AI tools reach classrooms. Each one uses a different definition of “safe.”

1EdTech TrustEd Apps

TrustEd Apps is the biggest player by volume. It lists more than 11,000 vetted products. Formal statewide partnerships make it the most embedded in state policy infrastructure. Vendor membership fees fund the program. Its review, however, covers privacy policy compliance only.

Common Sense Privacy Seal

The Common Sense Privacy Seal launched in 2025 and takes the opposite approach. Analysts use a rubric with more than 200 points. They also run independent reviews and quarterly monitoring. Fewer than 10 percent of evaluated products earn the seal. Philanthropy funds this work. Still, it only evaluates privacy, not whether AI tools actually work in class.

Digital Promise Responsibly Designed AI

Digital Promise also launched its certification in 2025, and it goes further. Vendors must show real data privacy practices. They must also work actively to reduce algorithmic bias. Transparency is required, so vendors explain how the AI makes decisions. Educators must be able to override those recommendations.

Thirty-nine district leaders, researchers, and product teams co-designed the framework. Chicago Public Schools and Denver Public Schools now build it into procurement. Independent funding keeps it free for districts.

Student Data Privacy Consortium

The SDPC takes a fourth path entirely. It is free, state-based, and focused on legal data-sharing agreements. But it does not certify AI tools at all. It simply makes the paperwork manageable between districts and vendors.

Four organizations, four definitions of safe. None of them is wrong, exactly. Yet measured against the full definition of safety, none does the whole job. Privacy, accuracy, learning, equity, and teacher control all matter together.

What States Should Do About AI Tools

States can close this gap with four practical steps.

Require evidence of effectiveness and equity

Privacy compliance alone is not enough. States that point districts to a vetting program should demand more. That program should show whether a tool improves learning for all students. It should also flag bias and accessibility barriers. The Digital Promise framework is a good model here.

Make vetting public and free

North Carolina’s GenAI Rubric for School Districts gets this right. Colorado’s AI Resource Evaluation Tool does too. Both are publicly available at no cost. No membership fee should block a rural Title I district from protecting its students.

Move from guidance to accountability

As of 2026, only a handful of states have binding requirements for district AI policies. These include Ohio, Tennessee, Idaho, Oklahoma, Maryland, and Virginia. Everyone else has published advisory documents. But guidance without accountability is just a website. Read more in our education policy coverage.

Designate someone responsible

Sometimes nobody at a state education agency owns AI governance. Then the job gets outsourced by default. It goes to a membership consortium, a vendor, or a teacher’s weekend research. Maryland’s new AI Ready Schools Act offers a better model. It requires every district to name a local AI coordinator.

Questions Worth Sitting With

Are you a state education leader, district administrator, or policymaker? Then ask yourself these four questions.

• Do you know which AI tools run in your schools today, including ones teachers adopted alone?
• Does your approval process check effectiveness and equity, or only privacy compliance?
• If a third-party seal approves your AI tools, do you know who funds that organization?
• Who in your system is accountable for those answers?

These are not gotcha questions. They are basic governance questions, and AI adoption has outrun them. Researchers, policymakers, and educators need to work on them together. Single scores rarely tell the full story, as we showed in our piece on how states rate schools.

Where Agora Goes Next

At Agora Research Academy, we are digging into these questions. An upcoming course on AI in education will map the national and state-by-state landscape. It will examine how schools adopt and fund AI tools. It will also ask what rigorous, equitable governance could look like.

In the meantime, we want to hear from you. How is your state or district vetting AI tools right now? Tell us what works and what does not.

Want more on this topic? Explore our A.I. articles or subscribe for new insights. You can also contact our team about research support.

FAQ,s