Submission 55
Alannah & Madeline Foundation
Submitter group: Civil society.
An AI model identified the submitter group and wrote the summary, key points and recommendation summaries. Counts do not measure importance or merit.
Summary
Read the original submission for the full text.
The Foundation argues children’s best interests should guide AI regulation, while recognising benefits children report. It recommends consumer protections, an overarching Digital Duty of Care, privacy enforcement, comprehensive digital literacy and AI safety research. Its supported attachment adds detailed legal reforms, research powers and children’s participation, treating literacy as a last defence after regulation and safety by design.
Key points
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.
Argues that the best interests of the child should be a priority principle in AI regulation.
We call upon the Committee to recognise the best interests of the child as a priority principle for the regulation of artificial intelligence
Argues that consumer law protections should recognise children as users of AI-enabled products even when they do not purchase them.
We see significant potential for consumer protections to lift AI standards for children, provided children are recognised as users of products and services, even when they do not purchase them personally.
Argues that a single overarching Digital Duty of Care with system-wide risk assessment is preferred over one focused on a few specific problems.
As risks associated with AI are evolving rapidly, we would caution against a Digital Duty of Care focusing too heavily on addressing a few specific problems. An overarching requirement on platforms to undertake system- wide risk assessment and mitigation seems more appropriate.
Says it supported the draft Children’s Online Privacy Code and expects an AI-inclusive final Code and resources to implement children’s rights protections.
We strongly supported the draft Children’s Online Privacy Code released for comment by the Office of the Australian Information Commissioner. We trust that the final Code will be AI-inclusive and that the regulator will be resourced appropriately to ensure its protections for children’s rights are brought to life.
The attached paper argues digital literacy should be a last line of defence after regulation and safety by design, rather than a substitute for regulatory action.
A focus on digital literacy however, needs to reflect the learnings from the past, and ensure that ‘digital literacy’ is not deployed as an intervention that avoids the need for strong regulatory action. Digital literacy skills can function only as the ‘last line of defence’ after regulatory reforms have been implemented and safety-by-design techniques adopted.
Recommendations
Recommendations are shown beside the submission’s own words. Listed in the same order as on its page in the interactive guide.
Recognise the best interests of the child as a priority principle for the regulation of AI.
We call upon the Committee to recognise the best interests of the child as a priority principle for the regulation of artificial intelligence (AI).
A program corrected obvious text-recognition errors in this quote for display.
Recognise children as users of AI products and services for consumer protection, even when they do not purchase them personally.
We see significant potential for consumer protections to lift AI standards for children, provided children are recognised as users of products and services, even when they do not purchase them personally.
Provide high-quality guidance on child safety standards for AI-enabled goods and services.
Introduce consumer protections for AI-enabled goods and services, supported by a Digital Ombuds service and high-quality guidance on child safety standards.
A program corrected obvious text-recognition errors in this quote for display.
Require platforms to assess and mitigate risks system-wide under an overarching Digital Duty of Care, rather than focusing on a few specific problems.
As risks associated with AI are evolving rapidly, we would caution against a Digital Duty of Care focusing too heavily on addressing a few specific problems. An overarching requirement on platforms to undertake system- wide risk assessment and mitigation seems more appropriate.
Resource the regulator appropriately if the Online Safety Act 2021 is strengthened to better regulate AI.
Any strengthening of the Online Safety Act 2021 to better regulate AI must be accompanied by appropriate resourcing for the regulator to ensure platforms are held accountable for their actions.
Empower and resource the OAIC to uphold children's rights under the Privacy Act 1988 in relation to AI.
Ensure the Office of the Australian Information Commissioner is empowered and resourced to uphold children’s rights under the Privacy Act 1988 in relation to AI.
Ensure the final Children’s Online Privacy Code includes AI and resource its implementation.
We trust that the final Code will be AI-inclusive and that the regulator will be resourced appropriately to ensure its protections for children’s rights are brought to life.
Prioritise comprehensive, AI-inclusive digital literacy for children and young people and their parents, carers and school communities.
Prioritise comprehensive, AI-inclusive digital literacy for children and young people and their parents, carers and school communities.
Make digital literacy encompass critical thinking about AI design and functioning and safe, responsible online decisions.
Digital literacy must involve more than technical skills. Children and young people are facing dilemmas ranging from the wellbeing implications of using AI companions to the ethics of sharing AI-generated imagery involving other children. They need support to build digital literacy which includes critical thinking about how technologies including AI are designed and function and how to make safe, responsible decisions online.
Align AI-inclusive digital literacy between home and classroom.
It is also important that AI-inclusive digital literacy is aligned between the home and the classroom, so that students are supported appropriately throughout.
Give particular digital-literacy attention to the most vulnerable children and young people, given their higher technology use, online risks and lower school engagement.
While digital literacy is vital for all children and young people, particular attention should be paid to the most vulnerable cohorts, who tend to show higher rates of digital technology use, higher rates of risk and harm online, and lower rates of school engagement than their peers.
Advance the science of AI safety to uphold children's best interests, including through the AI Safety Institute.
Advance the science of AI safety to uphold children's best interests, including through the AI Safety Institute.
Support legislation, regulation and policy with evidence on direct and indirect effects on children, emerging trends, capabilities, risks, harms and lessons from interventions.
Legislation, regulation and policy must be supported by high-quality evidence and analysis about how AI impacts directly and indirectly on children; emerging trends, capabilities, risks and harms; and outcomes and lessons from interventions by legislators, regulators, service providers, educators and industry.
Attached paper: Revise the Australian Consumer Law to introduce consumer safety standards and liability for AI goods and services.
A revision of the ACL to introduce consumer safety standards and liability for AI goods and services
Attached paper: Apply product-safety obligations to digital platforms acting as marketplaces for AI goods and services.
Product safety obligations, including for AI goods and services, on digital platforms that act as market places for these products (as intermediaries) would provide stronger protection for children.
Attached paper: Establish an adequately resourced and staffed Digital Ombuds service, or at least a Digital Ombuds for children.
The establishment of a Digital Ombuds service, or at the least, a Digital Ombuds for children. This would need to be adequately resourced and staffed
Attached paper: Review Child Safety Standards and their application to AI services and products operating in Australia.
A review of Child Safety Standards,32 and how they may apply to AI service and products operating in Australia
Attached paper: Provide ACCC guidance on unconscionable conduct in AI services and products for children, potentially informed by Child Safety Standards.
Guidance from the ACCC about what unconscionable conduct looks like on AI services and products for children, which could also be informed by Child Safety Standards
Attached paper: Use complementary regulation outside consumer law, potentially Online Safety Act reform, for transparency on Australian user numbers and AI risks and harms, plus researcher access.
Regulation that includes requirements for transparency, such as: ■ User data about the number of Australian users accessing a service or product ■ AI service or provider data about risks and harms ■ A researcher access scheme These would most likely be located outside of the ACL in other regulations, such as through reforms of the Online Safety Act
Attached paper: Use a strong Children’s Online Privacy Code and Privacy Act tranche II reforms to limit data sharing and prohibit targeting children, including the whole data-collection pipeline.
Privacy reforms — including a strong Children’s Online Privacy Code and tranche II of the Privacy Act reforms — that addresses and limits data sharing, and includes a prohibition on targeting children (including the whole pipeline of data collection)
Attached paper: Enable the Digital Platform Regulators group to fully implement and enforce existing rules on AI services and products.
Outside of the ACCC, the network of regulators associated with AI service and products — the DP Reg group — needs to be enabled to fully implement and enforce the regulations already in place
Attached paper: Replace “directed at children” requirements with “likely to be accessed by” requirements, with regulator guidance on the Australian standard.
Within the regulatory landscape, we need to adopt a regulatory approach that moves away from ‘directed at children’ requirements to ‘likely to be accessed by’ requirements. This will require guidance from regulators about what the likely to be accessed standard looks like in Australia
Attached paper: Consider stronger investigative powers or stronger use of existing powers to surface evidence and data about harms.
Stronger investigative powers for regulators, or potentially a more muscular approach to using the powers that they have, to surface evidence and data around harms
Attached paper: Revise the Online Safety Act for a singular, overarching Digital Duty of Care superseding BOSE and its codes, or at minimum explicitly extending beyond BOSE.
The Online Safety Act (OSA) needs to be revised, to include: ○ The introduction of a Digital Duty of Care (DDOC). The DDOC must supersede the BOSE and its codes, or at the very least, it must be made clear that the DDOC is not limited to the BOSE
Attached paper: Make that duty cover all systems, processes and current and future risks, including advertising, monetisation and content recommendation.
The DDOC must also be singular and overarching to ensure it covers all systems and processes that platforms deploy, and to cover all risks — both those we currently know the OSA cannot cover, and potential future unimaginable risks ■ Advertising systems and monetisation systems, as well as content recommender systems, must be covered by the DDOC
Attached paper: Replace industry classifications with all information services, or at least broaden them to AI services and products beyond chatbots.
The current industry classification scheme needs to be replaced with a broader focus on all information services or at least expanded to include AI services and products beyond chatbots.
Attached paper: Consider prescribed annual transparency reports on Australian child users and risks, researcher access and increased investigative powers.
Requirements for heightened transparency. This could include: ■ Proscribed annual transparency reports, including data such as the number of Australian child users and information about risk ■ A researcher access scheme ■ Increased investigative powers for regulators
Attached paper: Introduce licensing or onboarding requirements allowing added obligations and accountability for platforms and providers; include children’s participation in any licensing scheme.
A licensing scheme or onboarding requirements that allow additional requirements to be placed on platforms and providers, and increases accountability. Requirements for children’s participation should be part of any licensing scheme.
Attached paper: Include children’s data-removal rights and direct-marketing provisions in Privacy Act reform, and consider how it could address children’s data in AI training sets.
Reforms to the Privacy Act, which need to include; prohibiting targeting to children; the right to request data removal for children, and; direct marketing provisions (especially when it comes to ads for nudifying apps for example). In addition, consideration needs to be given to how the Privacy Act could address the use of children’s data in AI training sets overall
Attached paper: Explore Commonwealth and state criminal laws on sexting in relation to sexualised companion chatbots, including more effective enforcement or revision.
More effective enforcement of, or revision to, Commonwealth and State Criminal Codes, specifically when it comes to ‘spicy’ Companion AI Chatbots. There was a lack of clarity about if spicy discussions between Chatbots and children under 18 may invoke criminal codes around ‘sexting’.
Attached paper: Amend anti-discrimination law for algorithmic bias; explicitly consider intersecting disadvantage for First Nations children, children with disability, those in out-of-home care and those in low-income households.
Amendments to anti-discrimination laws to effectively address algorithmic bias (which is noted in passing in the National AI Plan). This also requires an explicit recognition that algorithmically mediated discrimination can compound existing disadvantages for children, especially for First Nations communities, children with disability, children in out-of-home care, and children in low-income households. Any review of anti- discrimination law in the context of AI should explicitly consider these intersections.
Attached paper: Undertake or commission an analysis of a holistic child-rights approach to AI products and services, covering participation, access and protection.
Undertaking or commissioning an analysis of what a holistic child-rights approach to AI products and services would look like, including participation and access rights as well as protection issues
Attached paper: Develop and use ISO-style industry standards and certification schemes.
The development and use of industry standards — in the International Organisation for Standards (ISO) model of Standards — as would certification schemes against these standards.
Attached paper: Consider onshoring requirements during Online Safety Act review to improve accountability.
The OSA review process also provides the opportunity to include requirements for licensing and ‘onshoring’ that could help drive up accountability processes,
Attached paper: Enable meaningful children’s participation across AI policy, extending beyond eSafety’s Youth Reference Group and encompassing education, health and other sectors.
Firstly, as a formal mechanism the Terms of Reference for the group and its members can have a limiting effect on public commentary, and secondly, there is a need for participation right across the ‘policy stack’ of AI governance debates — from school and university focussed policies to health care providers and so on.
Attached paper: Foster research on a child-rights framework for AI risks and capabilities, safe procurement in schools, child care and health care, and human security.
The research agenda emerging under the National AI Plan needs to foster research into: ○ A child-rights framework for understanding the risks and capabilities of AI products and services on children ○ Procurement of safe AI within schools, child-care and health-care settings ○ Human security as a safety issue
Attached paper: Research child-appropriate training sets in Australia and globally, and a future-focused sandbox for testing generative-AI guardrails and unforeseen risks.
Developing model training sets that are appropriate for children and built for children, both in Australia and globally ○ A future-focussed sandbox that allows researchers to test Generative AI guardrails, and other ‘unimaginable’ risks
Attached paper: Build participation and co-design into the research agenda, with civil-society support beyond an AI Safety Institute youth reference group.
Participation and co-design needs to be built into the research agenda. Civil society is skilled and able to support this, beyond developing a potential ‘youth reference group’ for the AI Safety Institute.
Attached paper: Support civil-society AI safety research outside the Institute through publicly available data, researcher access and different or enhanced funding models focused on children’s rights.
Support for civil society to meaningfully undertake AI safety research outside of the AI Safety Institute, including: ○ A researcher access scheme and requirements for publicly available data ○ Different or enhanced funding models for AI safety research as it affects children’s rights.
Attached paper: Give the AI Safety Institute investigative powers to test generative-AI production of child sexual abuse material, supported by a legal framework.
The AI Safety Institute should be given specific investigative powers to test the capacity of generative AI systems to produce child sexual abuse material. Without the ability to test guardrails, it is impossible to evaluate whether platforms are meeting their obligations. The Oxford Internet Institute and similar bodies provide a model for what this research capacity could look like. However, it requires a legal framework to support it.
Attached paper: Extend safety research across model training, safety mechanisms and input/output filtering, operational monitoring, and collaboration among regulators, industry and civil society.
Consideration of safety needs to extend across the technology life-cycle. This includes exploring risks and capabilities emerging from; the model training stage, to ensure children’s rights are respected as foundational models are developed; safety mechanism development and deployment, to explore for example how post-inference, input and output filtering function; monitoring systems to ensure adequate thought is given to how these models operate in practice and how they are used, and; collaboration between regulators, industry and civil society.
Attached paper: Consider youth-led and youth-consultative research within the wider research agenda.
Whatever the focus of the research, children’s right to participate needs to be respected. Youth-led research, as well as youth consultative research, has significant capacities to generate new insights and understandings, and needs to be considered as part of the broader research agenda.
Attached paper: Foster collaboration, capacity building and resourcing for independent research, including evaluation of how AI advances or threatens children’s rights.
There is a need for AI safety research that would sit outside the proposed AI Safety Institute. For example, civil society can develop and deliver impactful research, as can academia. Fostering this requires a deliberative approach to collaboration, capacity building and resourcing. For example: ● Transparency requirements around publicly documented risks assessments and data about risks and harms, and a researcher access scheme, could provide critical evidence to researchers ● Collaboration and capacity building around research for evaluation could help to develop a model for understanding how AI products and services advance or threaten children’s rights
Attached paper: Resource critical AI literacy, support evidence on what works, and treat literacy as the last defence after regulatory reform and safety by design.
Alongside research into safety, a comprehensive approach would require fostering enhanced critical AI literacy among children. Civil society may be well placed to develop and deliver such literacy schemes, with resourcing, in comparison to an AI Safety Institute. There is a strong role for research documenting evidence of ‘what works’ in digital literacy. A focus on digital literacy however, needs to reflect the learnings from the past, and ensure that ‘digital literacy’ is not deployed as an intervention that avoids the need for strong regulatory action. Digital literacy skills can function only as the ‘last line of defence’ after regulatory reforms have been implemented and safety-by-design techniques adopted.
Attached paper: Explicitly include children’s safety, privacy and participation in the AI Safety Institute’s foundational research and regulator-engagement mandate.
The establishment of an AI Safety Institute. The Institute presents an opportunity to embed children’s rights into its foundational work programme. Its mandate to generate technical research into AI risks and harms, and to engage with regulators, should explicitly include children’s safety, privacy and participation as core priorities from the outset.
Read the original
In some quotes a program corrected obvious text-recognition errors for display.