Child Safety Online

1,507 submissions in the snapshot (15 from organisations), portal total 1,507. Feedback list on the portal.

15 of 15 organisation submissions analysed; the model reads them as: opposes 6 wants changes 7 2 held back for review (low confidence).

• AI & Partners Company, NLD model reads as: wants changes source
  • Explicitly state children's rights under Article 24 of the Charter and UNCRC General Comment No. 25 in the text.

    Children's rights under Article 24 of the Charter and UNCRC General Comment No. 25 should be stated explicitly in the text.

  • Prohibit simulated intimacy and require crisis protocols for harmful content.

    Prohibit AI that: claims to be human; engages in romantic or sexual role-play; guilt-trips children or discourages human relationships; asks children to keep secrets.

  • Ban training models on children's conversations, emotional profiling, and in-chat advertising.

    Ban training models on children's conversations, emotional profiling and in-chat advertising.

• Electronic Frontier Foundation (EFF) NGO, USA model reads as: opposes source
  • Remove mandatory age gates for social media and video-sharing platforms that lead to privacy-intrusive age verification.

    1. mandatory age gates for social media and video-sharing platform, leading to privacy-intrusive age verification;

  • Reconsider safety by design measures that raise concerns about product-safety doctrines in speech regulation and federated services.

    2. certain safety by design measures, raising concerns about product-safety doctrines in speech regulation and their application to federated services; and

  • Address regulatory overlap with the EU's Digital Fairness Act agenda to avoid undermining consumer law.

    3. regulatory overlap with the EU's Digital Fairness Act agenda, risking undermining useful additions to the consumer law acquis.

• Faculty of Laws, University College London Academic or research institution, GBR model reads as: wants changes source
  • Address significant drafting errors and scope challenges arising from the reuse of Digital Markets Act definitions.

    Reuse of DMA definitions, which are designed for designation within, creates significant scope challenges.

  • Refine the overly broad risk criteria for Article 6 to avoid unintentionally capturing non-risky services.

    Risk criteria for Art 6 so broad to capture non-risky services, just because they are designed to incentivise interaction (as so many services are).

  • Clarify parental responsibility provisions to better account for complex family situations and domestic abuse risks.

    Parental responsibility is not held as digital data (e.g. court orders) for complex families, domestic abuse, etc.

• Micro company, POL Other, POL model reads as: opposes source
• Micro company, GBR Other, GBR model reads as: opposes source
  • Revoke the proposal and reconsider the approach.

    Please revoke this unfunny joke of a proposal, and come back with something with some actual thought and reason to it next time.

• Citizen Other, USA model reads as: opposes source
  • Do not pass the proposed act.

    Do not allow this act to be passed.

• Freie Spieleentwickler Deutschland Business association, DEU model reads as: opposes source
• Micro company, NLD Other, NLD model reads as: wants changes source
  • Recognise epistemic autonomy as an objective within online safety legislation.

    The EU KIDS Act should recognise epistemic autonomy: children need conditions in which they can constitute their own perspectives and remain capable of correcting them.

  • Incorporate attention architecture into safety-by-design obligations.

    Safety-by-design should thus include the design of attention.

  • Ensure digital environments protect minors' ability to form perspectives autonomously and self-correct.

    I suggest explicitly recognising: Digital environments for minors should protect and strengthen their capacity for autonomous perspective formation, exposure to other perspectives, and subsequent self-correction.

• Micro company, NLD Company, NLD model reads as: wants changes source
  • Make the public age assessment infrastructure directly usable by any provider at minimal or no cost.

    The public infrastructure should be usable directly, at no or minimal cost, by any provider.

  • Implement interoperable age information at the browser and operating system level instead of per-service authentication.

    Interoperable, transparent, privacy-preserving age information at the browser and operating system level would protect children without locking users in.

  • Avoid making teenagers' access to online spaces strictly dependent on guardian engagement.

    A framework that makes access contingent on a guardian's engagement will exclude the young people who may need those spaces most.

• Micro company, GBR Company, GBR model reads as: wants changes source
  • Include inferred, predicted, or modelled signals under Article 10(2)(b) rather than only signals from the minor directly.

    Cover signals inferred, predicted or modelled from the behaviour of other users, not only signals taken from the minor.

  • Establish evaluation metrics on quality, safety, and mental health as binding limits or objectives under Article 10(2).

    The metrics on quality, safety and mental health should be objectives or binding limits of the optimisation, not only a report.

  • Require auditors to possess specific machine-learning expertise and ensure audits inventory all available signals and compare evaluation metrics against independent measures.

    Auditors should have expertise in the objectives, input signals and measurement of machine-learning systems.

• Sambhali Trust NGO, IND; register: Non-governmental organisations, platforms and networks and similar model reads as: wants changes source
  • Test age assurance and parental tools on shared phones while minimizing personal data collection.

    Please test age assurance and parental tools on phones used by several family members.

  • Provide explanations of safety settings and restrictions in simple, age-appropriate language with pictures.

    Explain safety settings and restrictions in simple, age-appropriate language with pictures, and make it easy to find how to report harmful contact or content.

  • Ensure children have a confidential and direct path to support where accounts are managed or supervised by parents.

    Where parents set up or supervise a child's account, a child facing harm at home may still need a clear and confidential way to reach trained support, without unnecessary disclosure of the child's circumstances.

• Max-Planck-Institut für Informatik Academic or research institution, DEU model reads as: opposes source
  • Extend safety-by-design obligations and prohibitions on addictive design to cover all users.

    On a more positive note, the general obligation on safety by design and prohibition on intentionally addictive design are welcome, and should ideally be extended to cover all users of online platforms and services.

• Micro company, GBR Other, GBR model reads as: wants changes source
  • Target legislation to platforms and practices with demonstrable risks rather than broad portions of the internet.

    It should focus on platforms and practices that present demonstrable risks to children rather than imposing increasingly broad requirements across the wider internet.

  • Prevent age assurance from functioning as widespread identity collection and surveillance for adult users.

    Age assurance must not become a backdoor mechanism for widespread identity collection and surveillance.

  • Avoid creating a regulatory framework that compromises privacy, competition, innovation, and lawful internet access.

    Protect children, yes. But do not create a regulatory framework so broad that it compromises privacy, competition, innovation and lawful access to the internet for everyone else.