Brisk Teaching and Student Data Privacy: What Teachers Should Know
Brisk Teaching is a Chrome and Edge browser extension that embeds AI tools directly into Google Docs, Slides, and Classroom, and it only reads page content when a teacher actively triggers a specific task, not passively in the background. The company states it is FERPA and COPPA compliant, has signed the Student Data Privacy Pledge, and earned a 93% rating from the Common Sense Privacy Program, one of the higher scores among AI tools built for education.
None of that is a government certification. It is a mix of self-attestation and third-party review, and a district still needs its own signed agreement before a full rollout.
Quick Answer: Brisk Teaching accesses browser content only when a teacher clicks its icon and selects a tool, not continuously. It does not train AI models on submitted student work and states it does not retain the underlying text of a student's essay once a task completes. It has a real, independently reviewed privacy score (Common Sense Privacy, 93%), but no public evidence shows it belongs to an FTC-approved COPPA Safe Harbor program, so a signed district Data Processing Agreement still matters more than the marketing page.
Say it's the first week of a new grading period and a colleague mentions installing "this Brisk thing" that grades essays and builds quizzes without leaving Google Docs. Before rostering an entire class into a new browser extension, it helps to know exactly what it touches, what the company claims, and what still needs verifying. This guide walks through all three.
What Is Brisk Teaching, Exactly?
Brisk Teaching is a browser extension, not a standalone platform, meaning it layers AI-powered tools on top of apps a teacher already uses rather than replacing them. Understanding that architecture matters for privacy, since it changes what data the tool touches and when.
Founded in 2023 by Arman Jaffer, a former Chan Zuckerberg Initiative product lead, the company is headquartered in Los Altos, California. It raised a $5 million seed round in September 2024 led by Owl Ventures, followed by a $15 million Series A in March 2025 led by Bessemer Venture Partners (TechCrunch, 2025). By early 2025, the company reported more than 2,000 schools across 100 countries using its products, a figure it self-reports rather than one independently audited (TechCrunch, 2025).
The extension adds more than 30 AI-powered tools inside the browser, most commonly used inside:
- Google Docs, for writing feedback, rubric-based scoring, and tone or reading-level adjustments.
- Google Slides, for generating presentation outlines from existing lesson content.
- Google Classroom, for rubric-assisted grading through a feature called Brisk Grader.
- Any webpage or PDF, for turning existing text into a quiz, summary, or discussion questions.
Two features go further than a typical grading assistant. Inspect Writing replays a student's Google Docs edit history as a video, showing keystrokes, pastes, and deletions over time. Brisk Boost is the platform's only genuinely student-facing feature: teacher-configured AI chatbots students interact with directly for tutoring, debate practice, or exit tickets. Both are covered in more depth in Is Brisk Teaching Safe for Students?, which focuses on the safety and content-appropriateness side of those two features rather than the data-handling side covered here.
What Data Can Brisk Actually Access as a Browser Extension?
Installing any browser extension in Chrome triggers a broad-sounding permission warning, and Brisk's is no exception: it can technically read and change data on any site you visit. That warning describes what the browser's permission model allows, not what the extension actually does during normal use, and the two are worth separating carefully.
The Difference Between "Can Access" and "Does Access"
According to Brisk's own published explanation of its architecture, the extension behaves in three distinct stages:
- At install, it accesses nothing. The extension sits inactive until a teacher clicks its icon.
- On click, it checks page structure only, such as whether visible text is selectable, so it can decide which tools to offer. It does not read the page's actual content at this stage.
- Only when a specific tool is selected does Brisk process the relevant content, for example pulling the text of an open Google Doc to generate feedback on it.
| Extension state | What Brisk can see | What it actually collects |
|---|---|---|
| Installed, not opened | Nothing | Nothing |
| Icon clicked, menu open | Basic page structure (is text selectable) | Nothing from the page content itself |
| A specific tool run (e.g., Quiz Maker) | The content needed for that task | Only the content submitted for that task |
This trigger-based model is the practical answer to "what can Brisk see in my class's Google Drive." It cannot browse a student's entire Drive or read documents nobody has opened Brisk against, though it retains the same broad technical permission any full-page Chrome extension needs to function at all.
What Happens to Content After a Task Runs
Brisk's published security documentation states it does not use submitted input to train any AI models. Separately, it states it retains records of a request and the output generated, such as a feedback summary, but does not store the underlying student work itself, meaning a raw essay a student wrote is not kept in Brisk's systems after the feedback task completes.
That distinction, output retained versus source text discarded, is a genuinely specific claim, not a vague privacy-policy platitude, which is why it is worth confirming it appears the same way in your district's actual signed agreement rather than assuming the public page and the contract always match.
What Brisk States About FERPA and COPPA
Brisk describes itself as operating under FERPA's "school official" designation and states that schools provide the COPPA consent needed to collect data from students under 13. As the complete FERPA and COPPA guide explains in depth, neither law issues a certificate to a vendor, so every claim on this list needs to be read for what kind of evidence it actually is.
| Claim | What it actually is | Strength as evidence |
|---|---|---|
| "FERPA and COPPA compliant" | Self-description of Brisk's own practices | Weakest on its own; the norm across the AI-tool market |
| Signed a Student Data Privacy Pledge | A voluntary industry pledge, not a legal requirement or audit | Modest; signals intent, not verified practice |
| National Data Privacy Agreement with roughly two dozen states, including California, Illinois, Massachusetts, and Virginia | Pre-negotiated contract terms a district in one of those states can adopt | Enforceable, but only where signed |
| Common Sense Privacy Program rating of 93% | Independent third-party evaluation against a 200-plus question rubric | Strong; the most substantive outside signal on this list |
| 1EdTech TrustEd Apps Seal and Data Privacy Certification | Assessed certification from an education-technology standards consortium | Real and third-party reviewed, though specific to 1EdTech's own criteria, not FERPA or COPPA directly |
| No confirmed FTC-approved Safe Harbor membership (iKeepSafe, PRIVO, kidSAFE) | Absent from public materials reviewed for this article | The one gap worth asking about directly |
What the Common Sense Privacy Score Actually Breaks Down To
The overall 93% figure is an aggregate of nine separate category scores, and reading the breakdown tells a more precise story than the headline number alone.
| Category | Score | What it measures |
|---|---|---|
| Data Sharing | 100% | Protection against passing data to third parties |
| Data Security | 100% | Protection against unauthorized access |
| Data Safety | 100% | General responsible-use practices |
| Parental Consent | 100% | Protections specific to children's personal information |
| School Purpose | 100% | Following student-data-privacy laws specific to education |
| Ads & Tracking | 93% | Avoiding exploitation of user data for advertising |
| Data Rights | 88% | Whether users can access, correct, or delete their data |
| Data Collection | 83% | Clarity about what personal information is collected and why |
| Data Sold | 75% | Preventing outright sale of user data |
Read together, the pattern favors Brisk more evenly than a typical vendor evaluation, with every category above 75%. The comparatively lower Data Sold and Data Collection scores are still the two worth a closer look during your own review, even inside an otherwise strong result.
The Open Question: Which AI Model Actually Powers Brisk
Brisk's own materials describe using multiple large language models rather than a single provider, distinguishing a "Standard" tier available on the free plan from a faster "Turbo" tier on the paid plan. Public documentation reviewed for this article does not name the specific foundation-model companies behind either tier. That is not unusual among classroom AI vendors and does not by itself indicate a problem.
It is, however, exactly the kind of detail a district's technology office should ask Brisk to confirm in writing, the same way the pillar guide's checklist recommends for any AI tool: does a no-training commitment extend to whichever model actually processes the submitted text, not just to Brisk's own layer.
Setting Up Brisk for Your Classroom: A Practical Walkthrough
Getting Brisk running well, and privacy-consciously, from day one takes about fifteen minutes of setup before a first live lesson. These steps assume a school-provisioned rollout rather than an individual teacher's personal trial.
- Confirm your school or district has an account, provisioned through a school Google Workspace or Microsoft 365 login rather than a personal email, so usage falls under whatever agreement your school actually has in place.
- Install the extension from the Chrome Web Store, then sign in with your school-issued account rather than a personal Google account.
- Review the permission prompt Chrome shows on install. It will describe broad site access; that is a standard Chrome disclosure for any full-page extension, not a Brisk-specific red flag, but it is worth understanding rather than clicking past.
- Start with teacher-only tools first, such as feedback generation or quiz creation from a text you choose, before introducing Boost or any student-facing chatbot.
- If you plan to use Boost, decide on a join method in advance: a shared link, QR code, or class code lets students join without creating an account, which is the lower-exposure option for younger grades.
- Ask your technology office whether a district-level Data Processing Agreement or National Data Privacy Agreement exhibit has been signed, rather than assuming Brisk's public compliance claims automatically cover your specific school.
What Data Flows Through Each Core Feature
Not every Brisk tool touches the same category of data, and knowing which is which helps you decide what to roll out first. A feedback tool that only ever sees text you choose to submit carries a different exposure profile than a feature students interact with directly.
| Feature | What it touches | Student-facing? |
|---|---|---|
| Targeted Feedback (Google Docs) | The document text a teacher selects for feedback | No, teacher-initiated only |
| Quiz Maker | A webpage, video transcript, or document text a teacher provides | No, exports to Google Forms |
| Brisk Grader | Student submissions inside Google Classroom, plus a rubric | Indirectly; grades student work, but the teacher runs it |
| Inspect Writing | Google Docs edit history (keystrokes, pastes, timestamps) | Indirectly; reviews a student's process, not a live interaction |
| Brisk Boost | Whatever a student types into a configured chatbot activity | Yes, directly; the only feature students actively use themselves |
The practical takeaway: everything except Boost is something a teacher runs against content they chose, which keeps the data footprint closer to a normal grading workflow than to a student-facing chat product.
A Different Approach: Generating Material Instead of Scanning It
Brisk's tools mostly work downstream of existing content: a document already open, a webpage already published, a submission already turned in. A teacher who wants to keep student data out of a browser extension entirely for a given assignment has another option worth knowing about.
EduGenius can generate a differentiated worksheet, quiz, or writing prompt directly from a class profile set up with grade level and subject, with no live document or student submission ever needing to touch the tool in the first place. That's a different workflow than a browser extension layered on top of Google Docs, better suited to building an assignment from scratch than to giving feedback on one already written, and worth having as an option for the parts of a unit where minimizing what touches student work matters most.
A Grade 6 Rollout Scenario, Walked Through
Say you teach Grade 6 English and want to pilot Brisk for feedback on a persuasive-essay unit before deciding whether to introduce Boost later in the year. A cautious first month might look like this.
- Use Targeted Feedback on drafts you've already collected, rather than asking students to write directly inside a Brisk-connected session, keeping the rollout entirely teacher-initiated at first.
- Skip Inspect Writing during the pilot unless a specific concern comes up, since it is a heavier tool better introduced with a clear policy already in place, not tried casually mid-unit.
- Hold off on Boost until your technology office confirms the district's agreement covers student-facing use, since that is the one feature where students interact with the tool directly rather than a teacher reviewing output afterward.
- Tell students and families what the pilot involves, in plain language, before it starts rather than after a parent asks.
Pro Tips for a Privacy-Conscious Rollout
- Treat the free tier and a school-provisioned Pro account differently. A personal free account sits outside any district agreement by definition, which is fine for a short personal trial but not for a full class of student work.
- Ask specifically whether your district's National Data Privacy Agreement exhibit, if your state has one, has actually been executed, not just whether the state appears on Brisk's published list.
- Revisit the underlying-model question annually. The multi-model setup Brisk Teaching uses today is likely to keep evolving, and a policy set once at adoption can drift out of date within a year.
- Pair Brisk with a written classroom policy on Inspect Writing before turning it on, so a keystroke replay is used the way Brisk itself frames it: a conversation starter with a student, not an automatic verdict.
What to Avoid
- Rolling out Boost to a full class before confirming how student join data is configured. The link/code/QR join method needs no account by default, but a school can also enable Google or Microsoft sign-in to save progress, and those are meaningfully different exposure profiles.
- Assuming "FERPA and COPPA compliant" on a pricing page is the same as a signed agreement. It's a self-description, strengthened by real third-party signals like the Common Sense Privacy score and the 1EdTech seal, but still not a substitute for your district's own contract.
- Skipping the question of which foundation model powers Brisk's outputs. It is not named prominently in public materials, which makes it worth routing through your technology office rather than assuming an answer.
- Treating Inspect Writing's video replay as equivalent to an AI-detection score. It shows editing behavior, not a probability estimate, and Brisk itself recommends using it to start a conversation rather than to hand down a verdict.
Key Takeaways
- Brisk Teaching only accesses page content when a teacher actively selects a tool, not continuously in the background, despite the broad-sounding Chrome install warning.
- The company states it does not train AI models on submitted work and does not retain the underlying text of a student's essay once a feedback task completes, retaining only the request and generated output.
- Brisk earned a 93% rating from the Common Sense Privacy Program, built from nine category scores that all sit at 75% or higher, one of the stronger independently reviewed results among classroom AI tools.
- No public evidence shows Brisk belongs to an FTC-approved COPPA Safe Harbor program (iKeepSafe, PRIVO, or kidSAFE), which is worth asking about directly even alongside its other privacy signals.
- The specific foundation model or models behind Brisk's outputs aren't named in public materials, a detail worth confirming in writing before a district-wide rollout.
- Brisk Boost is the only feature where students interact directly; every other core tool is teacher-initiated against content the teacher selects.
- A National Data Privacy Agreement exhibit with your state doesn't roll out automatically; someone at your school or district still has to execute it.
Frequently Asked Questions
Is Brisk Teaching actually FERPA and COPPA compliant?
Brisk states it operates under FERPA's school-official framework and relies on schools to provide COPPA consent for students under 13. That is a self-description, not a government certification, since neither law issues one to vendors; a district's real protection is a signed agreement, and Brisk has also earned independent third-party signals like a 93% Common Sense Privacy rating.
What can Brisk see when it's installed in my browser?
Brisk accesses nothing when simply installed. Clicking its icon checks basic page structure only. It processes actual content, such as an open Google Doc, only when you select a specific tool to run against that content, not passively across every site you visit.
Does Brisk Teaching train its AI on student work?
No. Brisk's published security documentation states user input is not used to train any AI models, and that the underlying text of a student's essay is not retained after a feedback task completes, though the request and the generated output are kept.
Does Brisk Teaching have COPPA Safe Harbor certification?
No public evidence shows Brisk belongs to an FTC-approved Safe Harbor program such as iKeepSafe, PRIVO, or kidSAFE. Its compliance claim sits at the self-attestation and voluntary-pledge tier, strengthened by real third-party reviews like its Common Sense Privacy score and 1EdTech certification, but not a government-recognized COPPA audit.