The Pulse
OpenAI Pitches Six-Part Teen Safety Plan to Australia
OpenAI is asking Australia to adopt flexible, risk-based rules for teen AI safety through a six-pillar blueprint covering age assurance, harmful outputs, crisis response and parental controls.
AI.info Team ·
OpenAI is asking Australia to adopt flexible, risk-based rules for teen AI safety while Australian regulators push technology companies to take proactive responsibility for preventing harm. The company’s position arrives as the government develops a Digital Duty of Care framework that would require online services to identify and address risks before they reach users.
OpenAI published its Australian Youth Safety Blueprint on September 18, describing it as a policy contribution for protecting young people without cutting them off from tools used for learning, creativity and skills development. The document proposes six pillars, including age assurance, restrictions on harmful outputs, crisis-response procedures and parental controls.
The company says responsibility should not fall mainly on teenagers or their families.
“Companies must continue to build protections into their products from the outset.”
OpenAI says parents and guardians should receive accessible controls and clear information.
OpenAI Wants Rules That Adapt to AI Services
OpenAI supports an Australian legislative framework that requires companies to identify, assess and reduce serious risks to teenagers. It argues that lawmakers should focus on safety outcomes rather than prescribe fixed technical measures that could become outdated as products change.
The proposal sits alongside Australia’s planned Digital Duty of Care reforms. The country’s eSafety Commissioner describes that approach as placing the onus on digital platforms to proactively keep Australians safe and prevent online harms. OpenAI’s blueprint supports the objective but asks for obligations tailored to each service’s functions, users and risks, including the distinct features of generative AI.
Australia already regulates parts of the online ecosystem through eSafety codes and standards. The commissioner says registered age-restricted material codes cover services including generative AI services and AI companion chatbots, with restrictions involving sexually explicit material, suicide, self-harm and other harmful content. OpenAI’s document says any new framework should build on those measures rather than create overlapping requirements.
The Six Measures Behind the Blueprint
The first pillar calls for responsible adoption of AI in education. OpenAI says teenagers should learn to use AI critically and transparently, with teachers setting the pace in schools. The company also argues that AI literacy should count as a safety measure because students need to understand both the technology’s abilities and its limitations.
The second pillar focuses on privacy-protective age assurance. OpenAI recommends risk-based age estimation that distinguishes users under 18 from adults while minimising the collection of sensitive data. The blueprint says operating systems and app stores could help establish age where possible.
OpenAI says its Australian service uses the same type of age-prediction approach as its global systems. When the company cannot estimate a user’s age with enough confidence, ChatGPT defaults to a safer under-18 experience. Users placed there by mistake can confirm their age through Persona, a third-party verification service; OpenAI says it does not see submitted selfies or identification, while Persona deletes verification data within seven days.
The third pillar calls for under-18 safety policies covering serious risks such as self-harm, exploitation, grooming, sexually exploitative content and other high-risk interactions. Providers should test safeguards before deployment, monitor them after release and publish plain-language descriptions of the protections and parental tools they offer.
OpenAI Targets Emotional Dependence and Harmful Outputs
The fourth pillar addresses manipulative or deceptive AI behaviour. OpenAI says systems should not encourage emotional overreliance, compulsive engagement or secrecy from parents and trusted adults. They should also avoid presenting themselves as conscious beings, romantic partners or human authorities.
The blueprint says products should guide teenagers toward offline support when they detect distress and should make clear whether the user is interacting with a machine or a person. The approach treats emotional dependency as a product-design risk, not only as a matter of content moderation.
The fifth pillar calls for documented crisis-response procedures informed by research and external experts. Proposed measures include notifying connected parents when a teenager expresses suicidal intent, directing users to Australian services such as Triple Zero, Lifeline, Kids Helpline and 13YARN, supporting specialist organisations, and encouraging breaks during long sessions.
The sixth pillar concerns parental controls. OpenAI says parents should be able to link their account with a teenager’s account, manage memory, chat history and location settings, receive alerts associated with possible self-harm, set quiet or study hours, and receive notice when a teenager changes a previously enabled safety or privacy setting.
ChatGPT for Teens Arrived in Australia in August
OpenAI says the policy blueprint reflects safeguards already being introduced in its products. In August, the company began rolling out ChatGPT for Teens in Australia as the default experience for users identified as 13 to 17. The product builds on parental controls, under-18 behaviour principles and age assurance, with additional guardrails intended for younger users.
The blueprint claims that nearly nine in ten teens who use ChatGPT turn to it during a given week for learning, information, skill-building or productivity. OpenAI uses that figure to argue against treating AI services as identical to social platforms designed primarily for passive consumption.
That distinction will matter as Australia decides how its online-safety rules apply to AI products. The country’s existing codes already address certain chatbot risks, while the proposed Digital Duty of Care could place broader obligations on providers to prevent harm through product design.
Independent Assessments Would Test the Claims
OpenAI’s proposal calls for independent assessments of how effectively companies implement teen safeguards and reduce identified risks. Providers would publish plain-language summaries of those assessments while withholding details that could help users bypass protections.
The measure would give regulators, parents and researchers a way to judge whether age checks, crisis procedures and parental controls work in practice rather than relying solely on companies’ descriptions of their systems. OpenAI says common standards should support comparable audits across jurisdictions.
The immediate next step is Australia’s policy process. OpenAI says it is consulting with the government on the Digital Duty of Care approach and wants the final framework to preserve access to beneficial AI tools while imposing clear responsibilities on providers. Its six-part blueprint puts that balance on paper: access for education and creativity, with age-based protections and company accountability attached.