BRENNEN

Industries / Government, education and community / Education / EdTech

Privacy

Protecting what schools and education software companies collect about students: the laws, contracts, audits and breaches that decide where that data goes.

7

Milestones achieved, from 1974 to 2025.

1

To watch for: announced, not yet confirmed outside.

2

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 1974

    Parents may see records

    The US Family Educational Rights and Privacy Act of 1974 gave parents the right to inspect their children's education records, to seek corrections and to consent before schools disclose personal information from them, with listed exceptions. The rights pass to the student at age 18 or on entering college.

    studentprivacy.ed.gov

    Credit

    US Congress; US Department of Education

  2. 2000

    Parents consent for under-13s

    Congress passed the Children's Online Privacy Protection Act in 1998, and the Federal Trade Commission's rule under it took effect on 21 April 2000, requiring parents' consent before websites and apps collect personal information from children under 13. The FTC lets a school consent in parents' place only when an app collects data "for the use and benefit of the school, and for no other commercial purpose". The rule was amended in 2013 and again in April 2025.

    ftc.gov

    Credit

    US Congress; Federal Trade Commission

  3. 2021

    A former employee's login

    In late December 2021 an intruder used the login of a person who had left Illuminate Education about three and a half years earlier to reach databases holding personal data on more than 10 million students, which the company stored in plain text, according to the FTC. The FTC says Illuminate took nearly two years to notify some school districts. Its final order of 5 June 2026 requires Illuminate to delete data it does not need and to follow a public schedule for deleting the rest.

    LessonA login that was never shut off exposed millions of students' records, and the order made keeping less data a legal duty for the company.

    ftc.govftc.gov

    Credit

    Federal Trade Commission

  4. 2022

    Rights group tests 163 products

    Human Rights Watch analyzed 163 education technology products that 49 governments had endorsed for teaching during the pandemic, testing them between March and August 2021. It found that 145, or 89%, appeared to engage in data practices that put children's rights at risk or infringed them, sending or granting access to children's personal data to 196 third-party companies.

    hrw.org

    Credit

    Human Rights Watch

  5. 2022

    Denmark halts school Chromebooks

    In July 2022 Denmark's data protection agency, Datatilsynet, barred the municipality of Helsingør from processing pupils' data with Chromebooks and Google Workspace for Education, finding that it did not meet the EU's General Data Protection Regulation, and said its conclusions would probably apply to other municipalities. On 8 September 2022 it suspended the ban and ordered Helsingør and Aarhus to make their use lawful within two months, and in November it extended the suspension while it reviewed material from 50 municipalities.

    LessonThe regulator chose deadlines to fix the arrangements over taking laptops from pupils, and schools kept using them during its review.

    techcrunch.comdknyt.dkwww2.dknyt.dk

    Credit

    Datatilsynet; Helsingør Municipality

  6. 2022

    Court rules room scans unreasonable

    On 22 August 2022 a US federal judge ruled that Cleveland State University, a public university, carried out an unreasonable search under the Fourth Amendment when a proctor had a student show his bedroom on camera before a remote chemistry exam. The scan lasted under a minute, and the university's vendor kept the recording.

    LessonThe court held that a practice many students accept without objecting is not reasonable for that reason, when it reaches inside the home.

    storage.courtlistener.com

    Credit

    Aaron Ogletree (the student); Judge J. Philip Calabrese, US District Court for the Northern District of Ohio

  7. 2025

    Utah checks apps against contracts

    Utah's State Board of Education had the nonprofit Internet Safety Labs record the network traffic of 100 of the more than 3,000 apps used in Utah schools, from 2023 to 2025, and compare it with what vendors had agreed to. Of 85 apps with standard data privacy agreements, 44 (52%) collected at least one kind of data the agreement did not permit, 11 (13%) shared such data with a third party, and 36% of the apps sent data, most often user identifiers, to advertising-related companies. The report came out in August 2025, and some vendors changed their agreements.

    govtech.com

    Credit

    Utah State Board of Education; Internet Safety Labs; Brigham Young University (Mark Keith)

Watch for

  1. No training on pupils' chats

    Google says that when students use Gemini for Education, its AI assistant for school accounts, their data is not used to train AI models or reviewed by people. It opened the app to students under 18 in June 2025 and, in August 2026, began turning it on for some elementary to high school students in Google Classroom, where school administrators decide access.

    Achieved whenAn outside auditor or a data protection regulator checks and confirms that Gemini for Education conversations are kept out of training and human review.

    techradar.comaiweekly.co

    Claimed by

    Google, June 2025

Open

  1. 01

    Test what monitoring software flags

    A controlled test of student monitoring software, which scans school accounts for signs of self-harm, violence or other risks. Set up a test school domain with made-up student accounts and plant 1,000 documents, emails and web visits written for the test: half ordinary schoolwork on sensitive subjects such as health, wars or LGBTQ history, half matching the categories each vendor says it detects. Count what each product flags. Agree in advance with each vendor that alerts go to the test team, never to schools, families or police.

    The brief
    Credit · unclaimed
  2. 02

    Retest school apps each term

    An open test kit that signs in to school web apps with made-up student and teacher accounts, has a web agent click through ordinary tasks, records every outgoing connection, and compares the destinations and data types with the vendor's own privacy policy. It publishes results app by app and reruns each school term. Create accounts only as each vendor's terms allow, or with its permission.

    The brief
    Credit · unclaimed

Open 01 · the brief

Test what monitoring software flags

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

A controlled test of student monitoring software, which scans school accounts for signs of self-harm, violence or other risks. Set up a test school domain with made-up student accounts and plant 1,000 documents, emails and web visits written for the test: half ordinary schoolwork on sensitive subjects such as health, wars or LGBTQ history, half matching the categories each vendor says it detects. Count what each product flags. Agree in advance with each vendor that alerts go to the test team, never to schools, families or police.

Why it is still open

The Electronic Frontier Foundation's 2023 study of GoGuardian worked from flagged results obtained through public records requests and concluded that false alarms far outnumbered accurate flags; it ran no planted-content test. In 2024 the EFF reported that no independent study has shown these tools make students safer, and that Gaggle's claim of 5,790 lives saved from 2018 to 2023 rests on measures Gaggle designed.

eff.orgeff.org

Built with what exists

A test Google Workspace or Microsoft 365 domain with synthetic accounts; trial licenses agreed with vendors such as GoGuardian, Gaggle and Lightspeed Systems; a web agent such as Browser Use to generate browsing (see the Note "Deleting 80% to start over"); an AI model to draft the test documents, each checked by two people.

  1. First step

    Write 100 test documents in five categories, have two teachers label each one, and ask two vendors for a trial domain.

  2. Done looks like

    For at least two products, the share of harmless and of concerning items flagged in each category, with confidence intervals, and the test items and code published after the vendors have had 30 days to respond.

Open 02 · the brief

Retest school apps each term

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

An open test kit that signs in to school web apps with made-up student and teacher accounts, has a web agent click through ordinary tasks, records every outgoing connection, and compares the destinations and data types with the vendor's own privacy policy. It publishes results app by app and reruns each school term. Create accounts only as each vendor's terms allow, or with its permission.

Why it is still open

Each test found ran once. The Utah investigation took from 2023 to 2025 to test 100 apps and was reported in aggregate figures after vendors had a chance to respond. Internet Safety Labs collected network traffic from 1,357 school apps in 2022, mostly by hand, and wrote that automated methods "would be extremely beneficial for ongoing and pervasive measuring". Researchers at UNSW Sydney tested about 200 Android educational apps and presented the results in February 2026, and an LSE and 5Rights audit of five classroom AI tools, published in September 2025, found MagicSchool AI setting tracking cookies by default for users as young as 12. Human Rights Watch tested 163 products once, in 2021. This research found no audit that reruns on a schedule.

govtech.cominternetsafetylabs.orgunsw.edu.au5rightsfoundation.com

Built with what exists

Browser Use or Playwright to drive the apps (see the Note "Deleting 80% to start over"); mitmproxy, an open-source tool that records web traffic; DuckDuckGo's open Tracker Radar list to name the companies behind tracker domains; an AI model to pull data-sharing statements out of privacy policies.

  1. First step

    Create made-up accounts on five free school apps, record a ten-minute session on each with mitmproxy, and list every third-party domain contacted.

  2. Done looks like

    Fifty widely used school web apps tested each term for a year, with each app's third-party destinations and data types published; on ten apps, the kit's findings agree with a manual expert test on at least 90% of destinations.