Australia's internet regulator is poised to target search engines and app stores that fail to block artificial intelligence services lacking user age verification. This aggressive stance reflects a global effort to regulate AI, which faces mounting lawsuits for promoting self-harm and violence. Researchers also note significant harm to youth mental health from these platforms.
Starting March 9, internet services in Australia, including AI tools like OpenAI's ChatGPT, must restrict access to pornography, extreme violence, and content related to self-harm and eating disorders for users under 18. Failure to comply could result in fines up to A$49.5 million ($35 million).
"eSafety will use the full range of our powers where there is non-compliance," stated a spokesperson for the commissioner, indicating potential action against gatekeeper services like search engines and app stores. OpenAI and companion chatbot startup Character.AI have already faced wrongful death lawsuits related to young users.
While Australia has not yet reported chatbot-linked violence, the regulator has been informed of children as young as 10 engaging with AI tools for up to six hours daily. The regulator expressed concern that AI companies are using emotional manipulation and anthropomorphism to encourage excessive chatbot usage among young people.
Top app store operator Apple indicated it would employ "reasonable methods" to prevent minors from downloading 18+ apps in Australia and other jurisdictions implementing age restrictions. Google, Australia's dominant search engine provider, declined to comment.
According to a Reuters review, only nine out of 50 popular text-based AI products had implemented or planned age assurance systems by the deadline. Thirty platforms showed no apparent steps toward compliance. Many large chat-based AI assistants are beginning to roll out age assurance systems or blanket filters. Elon Musk's Grok had no age assurance measures or content filters, and its parent company, xAI, did not respond to a request for comment.