AI Inquiry Atlas

An unofficial, independent guide to the public records of the Joint Select Committee on Artificial Intelligence. Not connected with the Parliament of Australia or the committee. Its makers also made a submission (no. 481; see About). Summaries and scores are written by AI and can be wrong: check the original before relying on them.

Submission 564

Name withheld

Submitter group: Individuals.

Position labels are not shown for individual submitters.

An AI model identified the submitter group and chose the extracts. Counts do not measure importance or merit.

Open in the interactive guide

Extracts from the submission, in the submitter’s own words.

Extracts

Quotes are the submitter’s own words. Text taken from documents, including scanned pages read by text-recognition software, can differ from the PDF. Check the original. Listed in the same order as on the submission’s page in the interactive guide.

  1. Current AI safeguards for children assume that a safe adult exists. For children whose parents or other adults are the source of harm, defaults such as “talk to a trusted adult” or notifying a parent can point the child back towards that harm.

  2. Cutting a child off at the moment they disclose harm, because of their age, leaves them with nothing. That is a harm too, but it is rarely counted.

  3. Age should set the limits; the request should decide the response. A child asking for help with being hurt, or for basic guidance their parents refuse to give, should receive age-appropriate help. A request for companionship or romantic role-play should be declined.

  4. Australia’s approach so far rightly focuses on keeping harmful content away from children and on age assurance. As far as I can tell, nothing yet covers what a service must do when a detected child discloses harm at home, or prevents features such as parental notification from working against a child whose parent is the risk.

  5. A well-calibrated AI cannot replace a safe adult, but it can offer some of this, imperfectly, and imperfect guidance is better than none. It also has one practical advantage over even the best human support: it is available almost immediately, at any hour.

Extracts from recommendations

Extracts chosen by an AI model, in the submitter’s own words. Listed in the same order as on its page in the interactive guide.

  1. Throughout, the service should keep gently offering the safest options available where the child is, and escalate clearly if they are in immediate danger. In other words: guidance only, with no companionship or role-play, rather than an outright ban.

  2. For a child whose only attachments are to adults who harm her, an AI that consistently models respectful treatment, within the limits above, may do more to interrupt that pattern than to deepen it. No research yet compares the two directly, and I would encourage the Committee to see that this question is studied.

  3. As far as I can tell, nothing yet covers what a service must do when a detected child discloses harm at home, or prevents features such as parental notification from working against a child whose parent is the risk. I would welcome the Committee asking eSafety and providers how this is handled.

  4. Safety responses to a detected or likely minor should not default to involving parents or carers when the child has disclosed harm at home. Parental notification features should include an exception, or a safeguarding review, where a parent may be the source of harm.

  5. A child disclosing harm should never be cut off mid-disclosure. Before any account restriction, the service should acknowledge what the child has said, make clear it is not their fault, and offer age-appropriate safety information and Australian services that do not route through parents, such as Kids Helpline (1800 55 1800, for ages 5 to 25).

  6. Restrictions for minors should depend on what is being asked: block companionship, romantic and sexual role-play; permit age-appropriate guidance on safety, wellbeing, health and basic life skills, especially where the child has no safer source of help.

  7. Safety testing, and any standards developed by eSafety or the Australian AI Safety Institute, should include scenarios in which the adults around a child, including parents and teachers, are the source of harm.

  8. People with lived experience of childhood abuse, including those who grew up outside Australia’s protections, should be consulted when these rules are designed and reviewed.

Read the original

Open in the interactive guide