Submission 219
Bravehearts
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.
Argues that AI regulation should place responsibility for preventing child sexual abuse and exploitation on developers, providers and platforms rather than children. Recommends enforceable child-safety obligations, impact assessments, Safety by Design, testing, coverage of synthetic material and rapid reporting and removal, while acknowledging AI's potential social benefits.
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 children and young people should not carry primary responsibility for preventing child sexual abuse and exploitation in digital environments.
Bravehearts’ approach is grounded in the principle that responsibility for preventing child sexual abuse and exploitation should not rest with children and young people. Children have a right to participate in, learn through and benefit from digital environments without carrying the primary burden of protecting themselves from abuse and exploitation.
Argues generative AI challenges regulation primarily responding to particular content, platforms or conduct after harm has occurred.
However, generative AI challenges regulatory approaches that primarily respond to particular types of content, platforms or conduct after harm has occurred.
Recommends developers assess and mitigate foreseeable child sexual abuse and exploitation risks before deployment and after significant model updates.
AI developers and providers should have an affirmative responsibility to identify, assess and mitigate foreseeable risks of their technologies facilitating child sexual abuse and exploitation. This should include child-safety risk assessments before deployment and following significant model updates.
Argues that sexually explicit deepfakes of children are a form of technology-facilitated abuse that can cause serious harm even where no sexual activity took place.
The creation of sexually explicit deepfakes involving children should be understood as a form of technology-facilitated sexual abuse. A child does not need to have participated in sexual activity for sexualised imagery using their face, body or identity to cause serious harm.
Argues that regulation should be technology-neutral and risk-based, with obligations linked to capabilities, foreseeable uses and likely harms.
More durable approaches to AI governance therefore emphasise technology- neutral and risk-based regulation, with regulatory obligations focused to the capabilities of a system, its reasonably foreseeable uses and misuses, and the likelihood and severity of resulting harms
Says generative AI can automate or simplify offending activities previously requiring technical expertise, specialist software or significant time.
Generative AI potentially changes the economics of offending. Activities that previously required technical expertise, specialist software or significant amounts of time can increasingly be automated or simplified.
Recommendations
Recommendations are shown beside the submission’s own words. Listed in the same order as on its page in the interactive guide.
Establish enforceable child-safety obligations for developers and providers of AI systems, relating to the capabilities and foreseeable risks of those systems.
Establish enforceable child-safety obligations for developers and providers of AI systems, both to the capabilities and foreseeable risks of those systems.
Require developers/providers to assess and mitigate foreseeable child sexual abuse/exploitation risks, including child-safety assessments before deployment and following significant model updates.
AI developers and providers should have an affirmative responsibility to identify, assess and mitigate foreseeable risks of their technologies facilitating child sexual abuse and exploitation. This should include child-safety risk assessments before deployment and following significant model updates.
Require child-safety and child-rights impact assessments for higher-risk AI systems during development and release.
Require child-safety and child-rights impact assessments for higher-risk AI systems during development and release of these tools.
Embed Safety by Design principles in AI regulation, including requirements to demonstrate reasonable preventative safeguards against child sexual abuse and exploitation.
Embed Safety by Design principles in AI regulation, including requirements to demonstrate reasonable preventative safeguards against child sexual abuse and exploitation.
Require developers to demonstrate child-safety consideration throughout training, development, testing, deployment, monitoring and modification.
Developers should demonstrate that child safety has been considered throughout the AI lifecycle: training, development, testing, deployment, monitoring and modification
Require systematic child sexual exploitation risk testing, including ongoing assessment of attempts to circumvent safeguards.
Require systematic testing for child sexual exploitation risks, including ongoing assessment of attempts to circumvent safeguards.
For higher-risk generative systems, require safeguards against abusive generation and circumvention, ongoing misuse monitoring, accessible reporting/escalation and transparency about risks and safeguard effectiveness.
For higher-risk generative systems, requirements should include specific testing for child sexual exploitation risks; robust safeguards against generation of abusive material; measures addressing circumvention of safeguards; ongoing monitoring for emerging misuse; accessible reporting and escalation mechanisms; and transparency regarding identified risks and the effectiveness of safeguards
Ensure Australian law comprehensively addresses AI-generated and AI-manipulated child sexual abuse and exploitation material, including sexually explicit deepfakes depicting identifiable children.
Ensure Australian law comprehensively addresses AI-generated and AI- manipulated child sexual abuse and exploitation material, including sexually explicit deepfakes depicting identifiable children.
For children or representatives, provide rapid accessible reporting, removal, redistribution prevention, support information, escalation and remedies where providers failed reasonable preventative action, through trauma-informed processes suited to children's developmental capacities.
Rapid, accessible mechanisms should enable children or their representatives to report AI-generated sexual imagery, request removal and assistance preventing redistribution, obtain information about available support, have reports appropriately escalated, and seek remedies where providers have failed to take reasonable preventative action. These processes should be trauma-informed and designed around the needs and developmental capacities of children rather than the organisational structures of technology companies.
Require meaningful transparency from AI providers on child-safety risks, including vulnerabilities, mitigation measures, incidents and emerging patterns of misuse.
Require meaningful transparency from AI providers regarding child-safety risks, including identified vulnerabilities, mitigation measures, incidents and emerging patterns of misuse.
Strengthen research and data collection on AI-enabled child sexual abuse and exploitation, including grooming, sexual extortion, synthetic imagery, conversational AI and offender adaptation.
Strengthen research and data collection concerning AI-enabled child sexual abuse and exploitation, including grooming, sexual extortion, synthetic imagery, conversational AI and offender adaptation.
Include children, young people and victim-survivors in AI governance and policy development, using trauma-informed and ethically appropriate participation mechanisms.
Include children, young people and victim-survivors in AI governance and policy development, using trauma-informed and ethically appropriate participation mechanisms.
Ensure regulatory frameworks are technology-neutral and future-facing, with obligations linked to identified and foreseeable risk rather than limited to currently available technologies.
Ensure regulatory frameworks are technology-neutral and future-facing, with obligations linked to identified and foreseeable risk rather than limited to currently available technologies.
Invest in prevention alongside enforcement, recognising that law enforcement and content removal alone cannot respond adequately to AI-enabled harms at scale.
Invest in prevention alongside enforcement, recognising that law enforcement and content removal alone cannot adequately respond to harms occurring at AI-enabled scale.
Establish clear accountability across the AI supply chain so responsibility for child safety cannot be avoided through fragmentation between model developers, platforms and downstream services.
Establish clear accountability across the AI supply chain, so responsibility for child safety cannot be avoided through fragmentation between model developers, platforms and downstream services.
Have providers demonstrate reasonable, proportionate mitigation when foreseeable child-safety risks are identified.
Where foreseeable risks are identified, providers should be required to demonstrate reasonable and proportionate mitigation measures.
Recognise distinct prevention concerns for entirely artificial child imagery, without restricting harm to images produced through physical abuse of real children.
Synthetic material involving entirely artificial children raises different prevention questions, including its potential role in offender communities, sexual scripts, normalisation of the sexualisation of children and possible pathways towards offending involving children (Ó Ciardha et al., 2026). Regulation therefore requires sufficient flexibility to recognise these distinctions without adopting an overly narrow conception of harm based solely upon whether a real child was physically abused in producing an image.
Investigate sexualised, coercive, manipulative or developmentally inappropriate interactions between AI companions/conversational systems and children alongside generated imagery.
AI systems themselves may also create risks where conversational or companion technologies engage children in sexualised, coercive, manipulative or developmentally inappropriate interactions. These risks require investigation alongside AI-generated imagery.
Examine the ecosystem of AI-enabled offending rather than regulating technologies in isolation.
Policy responses therefore need to examine the ecosystem of offending, rather than regulating individual technologies in isolation.
Assign clear responsibilities across AI providers, platforms and hosting services for rapid, child-centred reporting, removal and control of generated sexual imagery involving children.
Develop rapid, child-centred mechanisms for reporting, removal and control of generated sexual imagery involving children, with clear responsibilities across AI providers, platforms and hosting services.
Invest in primary prevention (safe design, high-risk limits, defaults, child-rights assessments, education), early intervention (detection, monitoring, reporting, rapid action), and response (removal, investigation, enforcement, survivor support, redistribution/revictimisation prevention).
A useful framework would consider intervention across three levels: 1. Primary prevention seeks to prevent abuse before it occurs through safe AI design, appropriate limits on high-risk capabilities, effective default safeguards, child-rights impact assessments and public education. 2. Early intervention identifies emerging risks and disrupts pathways towards harm through detection of attempts to generate abusive material, monitoring patterns of misuse, effective reporting systems and rapid intervention. 3. Response ensures effective action after harm has occurred, including removal, investigation, enforcement, victim-survivor support and mechanisms to prevent redistribution and revictimisation. Investment across all three levels is essential.
A program corrected obvious text-recognition errors in this quote for display.
Read the original
In some quotes a program corrected obvious text-recognition errors for display.