In February, OpenAI said it was exploring how to enable generations including gore and erotica in age-appropriate contexts. The company also seemed to echo xAI by saying it was embarking on a mission to seek the truth by allowing users to have “intellectual freedom”, primarily by not refusing to generate the things they asked for, even if others found them offensive.
In March, OpenAI made a change allowing ChatGPT to create images of public and political figures, hateful symbols and racial features.
eSafety Commissioner Julie Inman Grant said creators of companion AI chatbots would need to prove children are not able to generate innapropriate content.Credit: The Sydney Morning Herald
In Wednesday’s post, Altman also said that ChatGPT would get new abilities to converse with a more human-like personality, and claimed the company could mitigate potential harmful downsides.
“We made ChatGPT pretty restrictive to make sure we were being careful with mental health issues. We realise this made it less useful/enjoyable to many users who had no mental health problems,” he said in the post.
“Now that we have been able to mitigate the serious mental health issues and have new tools, we are going to be able to safely relax the restrictions in most cases.”
The company has previously attracted criticism for its attempts to inject personality or bond-forming into its chatbot, particularly when it comes to children interacting with it. In August, an American couple, whose 16-year-old son had died by suicide, sued OpenAI claiming ChatGPT had contributed to his decision to take his own life. The couple claimed the platform was intentionally designed to foster psychological dependency and that safeguards to detect mental distress were easily sidestepped.
Loading
Inman Grant said she had proactively registered enforceable industry codes to protect children from age-inappropriate content, including the “clear and present danger” posed by AI-driven companion chatbots.
“I do not want Australian children and young people serving as casualties of powerful technologies, thrust onto the market without guardrails and without regard for their safety and wellbeing,” she said.
“Under the codes, AI companion chatbots will need to implement appropriate age assurance measures to prevent children from accessing harmful material, such as sexually explicit conversations, unless they can show they have enough strong guardrails to protect kids from generating this material in the first place.”
This codes also cover explicit violence, suicidal ideation, self-harm and disordered eating. Breach of a direction to comply will carry penalties of up to $49.5 million.
A spokesperson for federal communications minister Anika Wells backed eSafety and said the new codes would support the government’s work to restrict access to sexualised content online.
“Under these codes, generative AI services that have the highest risk of enabling children to generate pornography, self-harm material and high-impact violent material must implement appropriate age assurance measures,” the spokesperson said.
“The Australian Government is committed to protecting children from inappropriate content online.”
Get news and reviews on technology, gadgets and gaming in our Technology newsletter every Friday. Sign up here.
