Most conversations about school AI guardrails start with students and cheating. A report released last week suggests we’ve been looking in the wrong place first. The biggest exposure isn’t a student with a chatbot. It’s every adult in the building with an inbox, a login and a new AI tool they found on their own.
The report put it in one line that every superintendent should read twice: “The most targeted system in K-12 today is the human.”
I’ve spent four years working with districts and states on AI. This is the part of the conversation that rarely makes it into the board presentation. It should.
Key takeaways
- A new national report finds that about half of K-12 schools and districts had a cyber incident in the past year, and only 33% have formal AI policies.
- The report’s own conclusion is that the most targeted system in K-12 is the human, not the hardware.
- School AI guardrails are a leadership responsibility, not an IT checklist: they decide what data goes where, which tools are allowed and how people verify what they receive.
- Strong guardrails build judgment across the whole staff. Rules alone can’t keep up with AI-enabled phishing and impersonation.
What’s happening
On October 27, the Center for Internet Security (CIS) and its Multi-State Information Sharing and Analysis Center (MS-ISAC), working with CoSN, released The State of K-12 Cybersecurity 2026. According to the CIS and CoSN announcement of the 2026 K-12 cybersecurity report, 51% of surveyed schools and districts reported a cyber incident in the past year, and 34% said an incident affected school operations. More than 11,600 incidents required escalation and response during the 2025-26 school year. AI use is widespread across K-12, yet only 33% of organizations have formal AI policies. Nearly 80% of respondents said they need more cybersecurity support, resources or guidance.
K-12 Dive’s coverage of the report, by Anna Merod, adds detail. The survey covered 426 K-12 cybersecurity and technology leaders. Eighty-seven percent said phishing or email compromises are among the top incidents affecting their schools. Fifty-six percent said their cybersecurity programs are underfunded, and 70% said budget issues forced them to delay projects, technology or staffing. The report warns that schools without clear guardrails may expose sensitive information, bring unapproved tools into their environments and become more vulnerable to increasingly convincing AI-enabled phishing and impersonation attacks. It calls for a culture that treats staff as the first line of defense rather than the weakest link.
The next day, October 28, Sen. Bill Cassidy, chair of the Senate HELP Committee, sent a letter to Educational Testing Service about the California student data incident. The letter describes an incident in which information about millions of current and former California students was accessed and copied, including names, dates of birth, home addresses, demographic information and disability status. Cassidy asked ETS when it first learned of the incident, what protections it had in place and how it is communicating with affected families, with responses due November 12.
Put those together and the story is clear. The threat is getting more convincing. The data is more sensitive than most staff realize. And the tools are multiplying faster than the policies.
Why school AI guardrails are a leadership issue
In Chapter 4 of Human Still Required, I make the case for letting AI handle the busywork through the SAFE framework: tasks that are Structured, Administrative, First-draft and Easily repeatable. But I’m just as clear about the guardrails that make that possible. No AI-written communication goes out without a human read. No sensitive student data goes into unapproved tools. Drafts are starting points, not finished products.
Those aren’t technology rules. They’re leadership commitments. And this report shows what happens when they’re missing.
Tools can help you move faster. They can also help you make mistakes faster.
Human Still Required
A teacher pasting an IEP summary into a free tool to save twenty minutes isn’t acting in bad faith. They’re trying to keep up. But when only a third of organizations have formal AI policies, that teacher is making a data decision alone that the system should have made with them.
Here’s where Chapter 3 matters. The instinct after a report like this is to lock everything down. Block the tools. Add another mandatory training. Send another warning email. Some of that is necessary. But I argue in the book that “Control does not scale.” You can’t block your way out of a phishing email that sounds exactly like your superintendent. You can’t write a rule for every new tool a staff member finds on a Tuesday night. What scales is judgment: people across the system who know what matters, know what to protect and feel safe saying, “This doesn’t look right.”
That’s exactly the culture the report is describing when it talks about treating staff as the first line of defense. It’s not a technical posture. It’s a trust posture.
The trust problem AI impersonation creates
There’s a deeper issue here that connects to Chapter 8. AI-enabled impersonation works because it borrows trust that leaders have already built. An email that sounds like the principal. A voice message that sounds like the business manager. A request that sounds urgent and familiar.
In Chapter 8, I write about why staff trust their leaders in the first place:
They trust leaders because the message sounds like it came from someone who knows them.
Human Still Required
That’s the vulnerability. When a machine can imitate how we sound, the relationship has to carry more weight, not less. Staff need to know how their leaders actually communicate, what a real request looks like and that it’s always acceptable to pick up the phone or walk down the hall and ask. Verification isn’t distrust. In an AI world, it’s how trust protects itself.
And leaders need to model it. If a principal says, “I got an odd request that looked like it came from the district office, so I called to check,” that one sentence teaches more than a slide deck.
What leaders can do now
- Find out where you stand. Ask your technology leader one question this week: Do we have a formal AI policy that names which tools are approved and what data can never go into them? If the answer is no, you’re in the majority, and that’s the problem.
- Name the data lines in plain language. Staff don’t need a 30-page policy to start. They need two or three clear lines, such as: no student names, IEP details or health information in tools the district hasn’t approved. Put those lines where people work, not just in a handbook.
- Make verification normal. Tell staff, out loud and in writing, that checking a suspicious request with a phone call or a quick visit is always the right move, even if the message appears to come from you. Then thank people publicly when they do it.
- Ask your vendors the hard questions. The questions in Sen. Cassidy’s letter make a useful starting checklist for your own contracts: What protections are in place? How quickly will you tell us about an incident? How will you communicate with families?
- Treat this as professional learning, not a compliance module. One annual click-through training won’t build the judgment this moment requires. Short, regular conversations in staff meetings about real examples will.
If your leadership team wants a structured way to work through questions like these, the Human Still Required leadership guide is a good place to start.
Frequently asked questions
What are school AI guardrails?
School AI guardrails are the clear agreements that govern how AI is used in a school or district: which tools are approved, what data can and cannot go into them and how people check AI output and AI-generated messages before acting. Good guardrails are short, understandable and owned by leadership, not just the IT department.
Is AI making school cybersecurity worse?
According to the CIS and MS-ISAC report released October 27, 2026, AI adds risk in two ways. Schools without clear guardrails may expose sensitive information through unapproved tools, and attackers are using AI to make phishing and impersonation more convincing. The report also notes that only 33% of organizations have formal AI policies.
Should our district just ban AI tools to stay safe?
Blocking clearly unsafe tools is reasonable, but bans alone won’t keep a district safe. AI-enabled phishing arrives in the same inbox as everything else, and new tools appear faster than any list can track. The more durable protection is a staff that understands the data lines and feels safe pausing to verify.
The human layer
Every new technology has asked schools the same question in a different form. Who do we trust, and how do we know? AI just raised the stakes and shortened the time we have to answer.
The good news is that the most targeted system in K-12 is also the one leaders can do the most about. You can’t patch people. You can lead them.
So here’s the question I’d bring to your next leadership meeting: If a convincing message that sounded exactly like you asked a staff member for student data tomorrow, what would they do, and how would you know?
Sources
- CIS, MS-ISAC, and CoSN Report: K-12 Schools Face Growing Cyber Threats Amid Resource Constraints and Rapid AI Adoption, Center for Internet Security, October 27, 2026
- Cybersecurity incidents hit over 50% of schools and districts in past year, K-12 Dive, October 27, 2026
- Letter from Sen. Bill Cassidy to Educational Testing Service, U.S. Senate HELP Committee, October 28, 2026