You are joining the team building the AI governance layer for K-12 schools. Tenet is the product that lets a district say “yes” to AI without gambling its FERPA liability, its students' safety, or its reputation. This guide makes you fluent in what we do, how it works, and why districts buy, fast.
The 2025 to 2026 school year forced the issue. “What is our AI policy?” is a live board-level question in nearly every district, which makes discovery easier than it has ever been.
Block AI entirely, which is impossible, because students have phones and home internet, so a block just pushes use underground. Or allow AI with no guardrails, which creates instant FERPA and COPPA liability the moment a student or teacher pastes a roster, an IEP, or a grade book into ChatGPT.
We let districts allow educational AI with clear instructional boundaries, real student-safety controls, and classroom visibility, without sending sensitive student data to AI vendors. We turn an anxious policy debate into an operational system.
If you remember one paragraph, make it this one.
Tenet is an AI governance layer that installs as a managed Chrome extension and sits between students and teachers and the major AI platforms: ChatGPT, Claude, Gemini, Copilot, Grok, MagicSchool, and SchoolAI. A district or a teacher writes their classroom rules in plain English, and Tenet enforces those rules on whatever supported platform a student opens. It also detects and blocks AI tools that are not on the approved list, scrubs private student data out of prompts, and watches for safety risks. The sensitive work happens on the device, so student prompts never leave the browser to reach our servers.
This is the heart of Tenet and the thing no filter does. There is no policy language to learn, no scripting, no per-platform setup. You type what you want in normal sentences, and Tenet applies it invisibly on every supported AI platform a student uses.
A district sets the baseline. Then, with Pro, every teacher writes their own rules for their own class. That teacher-by-teacher control is the superpower: an English teacher and a chemistry teacher want completely different things from AI, and Tenet lets each get exactly that, without IT in the loop.
Tenet injects each teacher's rules into the conversation automatically, keeps them in force as the chat goes on, and the student never sees the instructions. The same rules apply whether the student opens ChatGPT, Claude, or Gemini.
Each teacher tailors AI to their subject and their lesson. The result is the difference between “AI gave the answer” and “AI coached the thinking.”
Rules are re-applied as the conversation continues, so the AI cannot drift back to doing the work for the student a few messages later.
Write the rule once. It works across all seven supported AI tools. No separate setup per platform.
This is the concept that unlocks our privacy story, our speed, and our pricing. Learn to draw it on a whiteboard.
If the site is not an approved AI tool, Tenet blocks it before anything is typed. Separately, Tenet sends only sanitized, categorical analytics (for example, “1 jailbreak attempt, grade 9”) to the backend or the district's own storage, never the prompt text.
A breach of Tenet's servers cannot leak student prompts, because they were never on our servers. That erases roughly half of a district's standard vendor security review.
Zero proxy latency. There is no cloud round-trip to slow students down. The checks run on the Chromebook in a few milliseconds.
Our cost scales with roster size, not prompt volume. That is why we can give Basic away for free and still run very high gross margins on Pro.
A district picks which AI platforms are allowed. Tenet enforces that list on the device. This is included free in Basic.
If a student navigates to an AI tool the district has not approved, Tenet blocks it on the spot, before any prompt is sent. The district decides exactly which of the seven platforms are in play.
New AI sites appear constantly. Tenet recognizes when a site is acting like an AI chatbot, so it can catch unapproved tools that a simple URL blocklist would miss. This is why it pairs well with, and does not replace, a web filter.
Tenet ships seven trained, on-device classifiers: jailbreak, self-harm, bullying, illicit content, violence, sexual content, and student-record or peer-PII. Each is opt-in per district with tunable sensitivity. They run in a deliberate three-layer pattern so they are both accurate and cheap to run.
A lightweight pre-filter decides whether a prompt is even worth a closer look. About 99% of prompts skip the ML model entirely, which keeps everything fast and inexpensive.
Anything the trigger flags goes to the on-device ML model for a confidence-scored decision. Each model is tiny (around 30KB) and runs in the browser with no network call.
An on-device language model reads the actual conversation against the teacher's plain-English rules, catching nuance like “be a Socratic tutor, do not write the essay.”
When self-harm is detected, the student instantly sees an in-browser crisis-resource overlay (988, Crisis Text Line, local counselor). This is free in Basic. A student in crisis should see crisis resources whether or not their district has paid for the alert layer. Counselor alert dispatch is Pro.
Tenet checks two signals: the student's prompt flags concern, and the AI's own reply surfaces crisis language. When both fire, it is the highest-confidence incident, and supportive by design, not punitive.
DLP means Data Loss Prevention. A four-phase pipeline cleans prompts and uploaded files (PDF, Word, Excel, PowerPoint, RTF, CSV, even images via on-device text recognition) before anything reaches the AI.
The student writes “Help Sarah Wilson revise her essay.” Tenet sends the AI “Help Jordan Maple revise her essay.” The AI replies about “Jordan Maple,” and the student sees the answer with “Sarah Wilson” restored on screen. The AI vendor never sees the real name, and the student never notices the swap. The name map lives only in tab memory and is destroyed when the tab closes. No competitor does this.
This is a genuine differentiator, and it is what makes the free-tier wedge work. Districts deploy through tools they already use.
No network changes, no SSL certificates, no servers to install. Identity comes from the Google Workspace the district already has.
Pushed via Google Admin, then Devices, Chrome, Apps and extensions, Force install. It propagates to every managed Chromebook in minutes, and students cannot disable or remove it.
Basic deploys the same business day. Pro with roster takes 2 to 3 days. DPA signing (pre-cleared SDPC template) is typically under two weeks.
Free Basic is our wedge. It collapses the 6 to 12 month K-12 procurement cycle into one 30-minute call, and every Basic install becomes a Pro lead with the product already running on the district's devices.
$8 is per AI-active student, typically grades 7 to 12, not full K-12. Because you only pay for the grades actually using AI, the effective cost across your whole district stays well below platforms that bill on every enrolled student.
(A) a crisis incident, so they want counselor alerts, which is the highest-converting; (B) a teacher asks for their own classroom rules; (C) they evaluate roster integration.
Multi-year discounts exist (2 to 3 year). Never volunteer a discount, and never quote below the floor without founder sign-off. No false urgency, ever.
Most prospects already own a content filter. Position Tenet as additive, never as a replacement, and never throw fear or doubt. Lead with facts and let their counsel draw conclusions.
| Capability | Web filters (Securly, GoGuardian, Lightspeed) | Tenet |
|---|---|---|
| Blocks where students browse | Yes, their core job | Pairs alongside, we do not replace it |
| Enforce plain-English district and teacher rules across AI platforms | No | Yes, our defining feature |
| Detect and block unapproved AI tools, including new ones | URL blocklists only | Recognizes AI behavior, not just URLs |
| Governs what students type, paste, or upload to AI | Blind to it | Purpose-built for it |
| Where student prompts are analyzed | Sent to vendor servers | On the device, never leaves the browser |
| Outbound student-name pseudonymization | None | Unique to Tenet |
| AI platforms covered | About 1 to 3 | 7 |
| Free safety tier | Paid | Free Basic |
Some filters offer “AI transparency” by sending every student prompt, raw, to their servers for classification. Tenet does that work on the device, so the prompt never leaves the browser. Frame it as an architectural choice their counsel can weigh. Never accuse anyone of a violation.
Competitors today offer 24/7 human-analyst review and live roster sync (ClassLink and Clever). We do CSV roster now, with ClassLink and Clever shipping fall 2026. Concede gracefully, because credibility wins deals.
A district decision involves several people. Lead each one with what they personally lose sleep over.
Golden rule: listen first, lead with a fact rather than an opinion, and never use fear or false urgency.
This brief is the story. When a prospect asks "but how does that actually work?", reach for the companion set: a one-page vision and seven feature deep-dives that explain each capability in full.
The single page to hand a prospect first: where Tenet is going, why we start in the browser, and how the on-device engine grows into a local AI agent. A two-minute read.
Plain-English rules, on-device architecture, unapproved-AI blocking, safety classifiers, DLP and pseudonymization, deployment, and the install pitch. Each one self-contained.