BRENNEN

Industries / Government, education and community / Childcare

Keeping parents informed

Keeping parents informed is the flow of news from a childcare center to the family, from daily reports, photos and messages to fees and quality ratings, and keeping that information safe.

4

Milestones achieved, from 2014 to 2025.

1

To watch for: announced, not yet confirmed outside.

3

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 2014

    Texts that prompt reading

    Benjamin York and Susanna Loeb of Stanford University tested READY4K!, an eight-month program of text messages to parents of preschoolers with small, specific tips for reading and talking with their children. They reported that parents' involvement at home and at school rose by 0.15 to 0.29 standard deviations, a standard measure of effect size, and children's early literacy by about 0.11. An adapted version, Tips by Text, was then tested with 3,649 four- and five-year-olds in 109 schools in England from 2019 to 2021. The pandemic cost about 70% of the follow-up data, and the 753 children who remained showed no significant gain in reading-related or social and emotional skills.

    Lessona message program that worked in one place needs testing again before it spreads, and a trial that loses most of its data cannot settle the question.

    nber.orgebi.ac.uk

    Credit

    York and Loeb (NBER Working Paper 20659, 2014); England analysis by F. O'Reilly, S. Loeb, A. Farrell and G. Scerif (University of Oxford)

  2. 2022

    Nursery apps leak and track

    Researchers from AWARE7, the Max Planck Institute for Security and Privacy, the Institute for Internet Security and Ruhr University Bochum examined 42 Android apps that childcare centers use to send reports, photos and messages to parents, along with the cloud servers behind them. They found that the apps could expose children's data to attackers, that many sent data to third-party trackers, and that what the apps did often did not match their privacy policies. The Electronic Frontier Foundation (EFF), a digital rights group, found that 7 of the 10 apps it checked that said they shared no data did so anyway, and asked the US Federal Trade Commission to review the gaps in the law; it said little was fixed after it alerted the app makers.

    petsymposium.orgeff.org

    Credit

    Moritz Gruber, Christian Höfig, Maximilian Golla, Tobias Urban and Matteo Große-Kampmann (Proceedings on Privacy Enhancing Technologies, 2022); EFF

  3. 2024

    A comparison site few use

    Australia's government website StartingBlocks.gov.au, run by the national quality authority ACECQA, lets parents compare services by quality rating, fees and vacancies. In its childcare inquiry, Australia's competition regulator (ACCC) found that the site "is not widely known or used", that its information can be out of date or missing, and that without session lengths parents cannot compare fees; many parents chose instead by visits and word of mouth.

    Lessonpublished data informs parents only if it is current, comparable and found where parents already look.

    accc.gov.austartingblocks.gov.au

    Credit

    Australian Competition and Consumer Commission, final report dated December 2023 and released in January 2024

  4. 2025

    Nursery chain held to ransom

    In September 2025 a group calling itself Radiant stole the names, photos, home addresses and parents' contact details of about 8,000 children from Kido, a chain of London nurseries. Infosecurity Magazine reported that the data was reached through the nursery's account on Famly, a parent-communication app, and that Famly said its own systems were not breached. The group posted profiles of about 20 children, phoned parents to press Kido to pay a ransom of about GBP 600,000 in bitcoin, and then said it had deleted the data. The Metropolitan Police arrested two 17-year-olds in early October 2025.

    Lessonone account that holds every child's photo and address is enough for an attacker, so who can sign in to it, and how, matters as much as the app's own security.

    therecord.mediainfosecurity-magazine.com

    Credit

    reporting by The Record, Infosecurity Magazine and Computer Weekly; Metropolitan Police

Watch for

  1. AI writes each child's day

    illumine, a childcare software company, says its software tags children in classroom photos and drafts a personal daily story for each child, with built-in translation, which teachers then review and approve. It gives no accuracy or error rate and cites no outside test.

    Achieved whenan outside party reports, on staged material, how often children are tagged wrongly and how many drafts teachers had to correct.

    illumine.app

    Claimed by

    illumine, October 2026

Open

  1. 01

    Grade app translations for parents

    A public test set of 300 made-up messages from a center to parents (daily reports, accident notes, allergy and illness alerts, closure notices), each translated by professional translators into 10 languages common among immigrant families, and a scoring method that counts errors and marks those that could cause harm. Apply it to the translation built into parent apps and to general AI models.

    The brief
    Credit · unclaimed
  2. 02

    Testing AI daily reports

    A staged test of software that turns staff notes, voice memos and photos into daily reports. Adult volunteers act out a scripted nursery day with dolls standing in for children and made-up names, keeping a true log of meals, naps, diaper changes and incidents; each tool's reports are scored against that log.

    The brief
    Credit · unclaimed
  3. 03

    Re-audit the nursery apps

    A repeat of the 2022 audit of childcare apps, on Android and iOS: for the 30 most-installed apps, record their trackers, whether two-step sign-in is offered, whether a photo link opens without signing in, and whether behavior matches the privacy policy, using only test accounts filled with made-up children.

    The brief
    Credit · unclaimed

Open 01 · the brief

Grade app translations for parents

Credit · unclaimed

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

What to build

A public test set of 300 made-up messages from a center to parents (daily reports, accident notes, allergy and illness alerts, closure notices), each translated by professional translators into 10 languages common among immigrant families, and a scoring method that counts errors and marks those that could cause harm. Apply it to the translation built into parent apps and to general AI models.

Why it is still open

Parent apps now translate automatically: by brightwheel's October 2026 comparison, ClassDojo translates messages into more than 130 languages and Bloomz into 108, and Daily Connect lists 10 languages. No published test of how well any of them translates childcare messages was found. Minnesota school districts found Somali, Hmong and Karen translations poor and apologized to families, but counted no errors, and Education Week reported in September 2026 that US schools using AI translation rely on spot checks and back-translation, with no measured error rate. Hospitals have such tests: Kong and colleagues (2026) had professional translators back-translate 316 sentences of emergency discharge instructions and found that 16% to 66% of instruction sets had at least one error, depending on tool and language.

mybrightwheel.comgovtech.comedweek.orgebi.ac.uk

Built with what exists

Made-up messages written with early-years staff; professional translators; the apps' own translation on test accounts; ElevenLabs Scribe v3 to add spoken voice notes in the same languages (see the Note "Fewer word errors"); open models run locally with ggml-based tools (see the Note "A first stable number for local AI").

  1. First step

    Write 50 messages with two nursery staff, covering the most safety-critical message types, and have them translated into Spanish and Arabic.

  2. Done looks like

    300 messages in 10 languages; each machine translation scored by two bilingual raters; error and potential-harm rates per tool and language published with confidence intervals; the test set released for reuse.

Open 02 · the brief

Testing AI daily reports

Credit · unclaimed

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

What to build

A staged test of software that turns staff notes, voice memos and photos into daily reports. Adult volunteers act out a scripted nursery day with dolls standing in for children and made-up names, keeping a true log of meals, naps, diaper changes and incidents; each tool's reports are scored against that log.

Why it is still open

illumine (October 2026) and Daily Connect say their AI drafts daily reports, and Daily Connect also incident notes. Neither gives an error rate or cites an outside test, and a search found no published test of such tools.

illumine.appen.dailyconnect.com

Built with what exists

Dolls and a mock classroom; ElevenLabs Scribe v3 for staff voice memos (see the Note "Fewer word errors"); small open models that pull fields out of notes, such as Liquid AI's LFM3 (see the Note "One GPU to fine-tune") or Meta's Muse Ember (see the Note "The jobs agents repeat"); vendors' trial accounts.

  1. First step

    Write one scripted day with a nursery practitioner and run it through two tools on trial accounts.

  2. Done looks like

    10 scripted days, each with 12 made-up children and at least 60 logged events; for each tool, the share of events reported correctly, missed, or put on the wrong child, with scripts and scores published.

Open 03 · the brief

Re-audit the nursery apps

Credit · unclaimed

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

What to build

A repeat of the 2022 audit of childcare apps, on Android and iOS: for the 30 most-installed apps, record their trackers, whether two-step sign-in is offered, whether a photo link opens without signing in, and whether behavior matches the privacy policy, using only test accounts filled with made-up children.

Why it is still open

The 2022 study covered 42 Android apps, EFF said little was fixed after it alerted the makers, and the 2025 Kido theft went through a nursery's app account. Flaws keep turning up one product at a time: in January 2025 the security researcher Florian Hantke reported that a free test account on KigaRoo, German kindergarten software, could open other users' records by changing a number in the web address, exposing by his estimate about 846,000 child records, and the company fixed it over a weekend. Searches of Crossref and arXiv found no follow-up audit across apps.

petsymposium.orgeff.orginfosecurity-magazine.comnetzpolitik.org

Built with what exists

MobSF (Mobile Security Framework) for reading app packages; Exodus Privacy's tracker list; mitmproxy to watch the test accounts' own traffic; an AI model to compare each privacy policy with the traffic seen (see the Note "Seven more flagged"). No probing of the makers' servers, and findings go to each maker 90 days before publication.

  1. First step

    List the most-installed childcare apps in five countries and set up two test centers with made-up children on the first five apps.

  2. Done looks like

    A public table of 30 apps against 10 checks, each finding reproducible from published scripts, with a count of how many 2022 findings still hold.