Elon Musk's X Corp was asked in a survey what sort of "harmful experiences or content" digital platforms should stop Australian children from seeing.
"None," the social media titan said.
The federal infrastructure department's Sarah Vandenbroek told the Royal Commission on Antisemitism and Social Cohesion on Friday a survey was conducted in 2025 to gauge whether social media organisations felt they had a digital duty of care.
X ticked "none of the above" in response to a series of questions about whether it should take responsibility for stopping those younger than 18 from seeing various kinds of online content.
Ms Vandenbroek said while the response had been "disappointing", social media agencies had a financial incentive to keep graphic content on their feeds as it generally had high engagement and attracted advertising.
A mandated duty of care for social media platforms was at least 18 months away, Ms Vandenbroek said, despite a 2024 online safety law review recommending its implementation.
Her department was adequately staffed to put the duty of care in place, she said, but the "complexity" made it a long process.
"A lot of the work that was done when the report was first handed over was redone with the new office to make them aware of the recommendations and the content, which is quite extensive," she said.
Despite that, the department had not sought legal advice on the 2024 recommendation that major social media platforms be required to have a local presence for the purpose of facilitating enforcement, nor had it interrogated all safety elements of the report.
Ms Vandenbroek was told members of the Jewish community, including Special Envoy to Combat anti-Semitism Jillian Segal, were frustrated by the lengthy timeline to enact new safety regulations.
"I wasn't previously aware of the very strong views held in the community, so that is something we can certainly do further work on," she said.
Meanwhile, Anthropic's first Australian general manager Theo Hourmouzis said automated content classifiers, common in artificial intelligence models, were effective at detecting hate speech when it was generated.
It included intentional misspelling, emojis used for hidden meanings and coded anti-Semitism.
"I am not sure how they (social media companies such as Meta and X) do their content moderation," he said.
"Anthropic regards this as technically very straightforward."
Anthropic's flagship product, a series of large language models known as Claude, is in theory more ethical than other AI systems, as it is governed by a strict constitution.
The royal commission resumed with a public hearing block on Monday after weeks of evidence being given in private.
The current hearing block is examining the role of social and mainstream media in perpetuating anti-Semitism.
Various Jewish groups and technology companies gave evidence in the first week, with public broadcasters the ABC and SBS expected to appear during the second week.
