Blog
AI grading privacy is a checklist you can take to any tool before student work leaves your desk. Is AI grading safe for student data? Start with what is collected, whether papers are used for model training or advertising, which processors see the work, how long it stays, who can open the results, how papers are handed back, and your school or district policy.
The questions below are for any AI grading tool. After the checklist, this page states how Graider answers each item, using the facts on the privacy policy.
Graider is the AI red pen for paper grading, a product of DM Liquidity LLC. Teachers photograph the stack, grade on their own answer key and rubric, review and toggle feedback, keep the pen, and hand work back as a PDF or an email. Pasting student work into a general chatbot raises its own questions about names and writing leaving the desk. That situation is covered in ChatGPT grading.
Your key. Their papers. Comments you release. You keep the pen.
Ask for an inventory before you upload a roster or a stack of photos. Account data usually means a name and an email, plus how sign-in works. Class and grading data is the larger set: classes, tests, answer keys, rosters, paper photos, text read off those photos, marks, and feedback. Also ask about device tokens, purchase status, and logs used to run the service and fix errors.
Does the tool use student papers or account data to train its own models? Does it use that data for advertising? Does it sell personal information? Ask again for the company that reads the page and the company that produces the marks. If the privacy page is silent on training, advertising, or selling personal information, take that silence back to your school.
Grading a photographed paper can pass through several companies: sign-in, reading the page, producing marks, storing the image, hosting the app, and running the job in the background. Ask for each name and what that company receives. Ask about subprocessors too: if a model is reached through another service, name both the service and the models.
Ask how long account data stays, and how long grading data stays. Ask whether paper photos follow the same clock as the account, what deletion removes, where the control lives, and how soon copies of uploaded files leave storage.
Ask who can open a class, a roster, and a marked paper. Ask whether result sharing is a step you take, and what leaves the account when you take it. Ask which processors can see marks and feedback, and for which job.
Ask what the student receives. A marked PDF, an email, and a file you download are different copies of the same work. Ask whether you can turn comments off when you will go over the items in class. The teacher who photographed the stack should still decide what goes home. Quizzes and unit tests follow that loop on AI test grading: photograph the papers, mark against your key, review, then release.
The checklist tells you what to ask the tool. Your school or district policy tells you whether you may use it. Read that policy before a roster or a paper photo is uploaded. When the wording is unclear, ask the person who owns the policy, and bring the vendor's privacy page to that conversation.
Graider publishes these answers on the privacy policy. What is Graider describes the grading loop around them: photograph the stack, grade on your own rubric, review, and keep the pen.
What is collected. Sign-in uses Clerk, with email, Google, or Apple. Account data is the name and email from that provider. Class and grading data includes classes, tests, answer keys, rosters, paper photos you upload, OCR text, marks, and feedback. Device data is an optional push notification token so Graider can alert you when a grading job finishes. Subscription data is purchase status through RevenueCat. Graider does not store your full payment card details. Usage and diagnostics are basic logs needed to run the service and fix errors, for example API failures.
Training and advertising. Paper images and text are processed with AI models solely to produce marks and feedback against the answer key and rubric you supply. Graider does not use student papers or account data to train its own models, and Graider does not use them for advertising. Graider does not sell personal information.
Processors. Graider shares data only with processors that help run the service. They receive only what is needed for that job.
Retention and deletion. Graider keeps account and grading data while your account is active. You can delete your account in the app: Account, then Delete account. Deletion removes associated Graider data promptly. Copies of uploaded files in object storage are removed within 24 hours.
Access and handback. Marks and feedback are part of class and grading data. Result sharing is a feature you request. You review the marks, toggle feedback on or off, and hand work back as a PDF or an email. You keep the pen. Release waits until you send it.
School and district policy. Graider is designed for teachers. Student work may be uploaded by teachers. Graider does not knowingly market the service to children under 13. Teachers are responsible for using the product in line with their school's policies. Check your school's or district's policy before you upload a class set.
Ask what data is collected, whether student work is used for model training or advertising, which processors and subprocessors see it, how retention and deletion work, who can access results, how papers are handed back, and whether the tool fits your school or district policy.
It depends on the tool's written answers and on your school or district policy. Graider publishes its answers on the privacy policy at graider.org/privacy. Confirm the fit with your school or district before you upload student work.
Graider processes paper images and text with AI models solely to produce marks and feedback against the answer key and rubric you supply. Graider does not use student papers or account data to train its own models, and Graider does not use them for advertising. Graider does not sell personal information.
Clerk handles sign-in. Reducto reads photographed papers and does not train on Graider data. OpenRouter and the AI models called through it produce marks and feedback and do not train on Graider data. PostgreSQL stores records, Supabase Storage stores paper images, Vercel hosts the app, and Render and Redis run background jobs. RevenueCat handles purchases. Expo delivers push when you enable notifications. Each receives only what its job needs.
Graider keeps account and grading data while your account is active. Delete the account in the app under Account, then Delete account. Deletion removes associated Graider data promptly. Copies of uploaded files in object storage are removed within 24 hours.
Read the marks, toggle feedback on or off, then hand the papers back as a PDF or an email. You keep the pen.