50% of Australians Use AI, But 80% Want Locally-Tailored Models

For the AI community, these findings present a dual mandate: massive adoption coupled with deep distrust of offshore technology. While half the nation now uses tools like ChatGPT, the call for homegrown models that avoid 'white, male, American-centric values' is more than a cultural plea — it's a market signal for localized, responsible AI development in a country that could become a key test bed for sovereign AI.

A newly published report from the Australian National University in partnership with Google has crystallized the nation's escalating anxieties around artificial intelligence, particularly in the political sphere. The study, based on a survey of more than 3,500 Australian adults and in-depth interviews with senior organizational leaders, reveals that almost half of all Australian adults now use AI tools — a figure projected to skyrocket as adoption among young people surges. Yet this rapid uptake is shadowed by deep public mistrust: a striking eight in 10 Australians expressed concern about the growing use of generative AI in politics, citing fears of misinformation dissemination and the potential leakage of sensitive political data. These numbers are not just abstract polling data; they represent a clear mandate from the electorate for immediate regulatory, technical, and policy responses.

“

A newly published report from the Australian National University in partnership with Google has crystallized the nation's escalating anxieties around artificial intelligence, particularly in the political sphere.

The report comes at a critical juncture. Australia is heading toward a federal election within the next year, and the capacity of tools like ChatGPT, Google Gemini, Dall-E, and Claude to generate convincing deepfakes, synthetic media, and targeted propaganda has already been demonstrated globally. The lead author, ANU researcher Dr. Jessica Herrington, pointed to real-world examples of deepfakes and online misinformation, noting that many risks “cannot be fully mitigated,” but that awareness is a crucial first step. Her concern extends beyond the immediate threat of false information. She highlights the structural vulnerability of Australia's reliance on overseas AI models, particularly those developed in the United States. These models are trained predominantly on American web data, inherently embedding cultural biases — what she describes as “white, male, American-centric values.” This bias, she argues, makes the models not only potentially culturally insensitive but also less effective and fair when applied to Australia's multicultural society.