Classroom.ai

Data & privacy

Your learning belongs in your workspace.

What is saved

Classroom saves your goals, source text, submitted answers, saved drafts, tool workspaces, feedback, assessment evidence and reflections in server-side records tied to your workspace identity. Other learners cannot access those records through the app.

Guest entry uses a signed, HttpOnly cookie that expires after seven days. Sign in to move guest rooms into your account and resume on other devices. Classroom.ai uses email and password accounts with 30-day HttpOnly sessions. Passwords are stored as salted, secret-peppered hashes; session tokens and recovery codes are stored as hashes. Email is an unverified sign-in identifier. Password recovery uses the private code issued when you create an account, and signs out existing sessions. The earlier private Sites preview uses its hosting service’s separate sign-in system.

AI processing

Relevant source passages, the current question, your answer, recent conversation and saved workspace observations are sent from the server through the owner-configured AI gateway to the selected model for tutoring or assessment. This beta uses Grok 4.7. Requests include store:false, but that does not establish a zero-retention agreement with the gateway or its upstream services. Their retention and monitoring policies also apply. The provider credential stays on the server.

Uploaded material and execution

For supported files, the browser extracts text before upload. The original file is not retained by Classroom. PDF scans without selectable text require OCR first. Images, diagrams and visual equations are not interpreted. Extracted source text is stored in separate page-located chunks with import coverage and hashes. Wikipedia imports contact only supported official Wikipedia hosts and retain a revision, section metadata, attribution and a prose snapshot. When enabled, YouTube imports send only the canonical video URL and optional caption language to Supadata to retrieve existing captions. Your questions, answers and account cookies are not sent to that service. Public YouTube oEmbed metadata may provide the title and creator; it is not lesson evidence. Caption text, supplied timestamps, language, available creator metadata, snapshot hash and import time are saved privately. Original video/audio and signed caption-access URLs are not retained. The provider key remains server-side. Native-caption mode does not request AI-generated transcripts or automatic retries. Following a video/timestamp link opens YouTube and its policies apply. Your CSV files may be loaded into the isolated Python learning runtime on your device. The runtime downloads pinned Pyodide assets from jsDelivr; it cannot call Classroom’s account APIs.

Export and removal

Export a room’s brief or learning record from Session recap. Remove the room from Classroom settings to delete its material metadata, extracted text chunks, conversation, workspace and evidence. Removing selected material also clears the session derived from it. Daily service-usage counters are separate from room content and are not reset by room deletion.

Public-page analytics and service limits

Anonymous visits to the public landing page can increment an aggregate daily counter. The counter stores the day, the route “/” and a count. It does not receive your answers, documents, room URLs, query strings or account identity. Development visits are excluded, and Do Not Track, Global Privacy Control and the choice below are respected.

The public landing page also sends limited measurement pings to Google Analytics (property tag G-040RWZ80Z6) automatically, without an opt-in prompt. Analytics and advertising storage remain denied: no Analytics cookies are read or written by this integration. Advertising personalization and Google signals are disabled. We send a fixed public page URL. Source titles, URLs, filenames, questions and answers are excluded from analytics. We do not send account identifiers, goals, answers, uploaded material, room IDs, query strings or referrers. Google still receives network requests and associated technical information; cookieless does not mean no data is processed. The tag is disabled when you enter setup or a learning room, and it is not loaded for signed-in learners, private room links, local development, Do Not Track, Global Privacy Control or the opt-out below. See Google’s privacy policy.

Guest AI requests and sign-in attempts may use secret-keyed, time-bounded identifiers for abuse prevention. Raw network addresses are not stored in those counters. This service-limit mechanism is separate from public traffic statistics.

Return to my classroom