Discussions about AI are happening everywhere, from ASX100 board rooms to local council planning committees to primary school drop off zones. And they’re all needed. These technologies are entering everyday life fast, and the question of what kind of future we are building is no longer abstract. For us, the answer has to be feminist in principle, and in practice.
In December 2025, the federal government released its National AI Plan. It arrived with the right vocabulary: safety, responsibility, innovation. But vocabulary without architecture is just aspiration. The Plan will only matter if it is implemented in ways that address the harms AI is already causing which currently hit women hardest.
The first thing to understand about AI is that it’s not neutral. Technology never is.
Facebook began as a tool to rate women’s attractiveness. Men have been harassing women online since the internet’s earliest days. These are not edge cases, they are features of how digital systems have been designed, funded and governed. Grok has demonstrated the misogynistic outputs that follow when guardrails are treated as obstacles to “free expression”. The Guardian recently reported that a woman was denied psychiatric treatment because she refused to consent to an AI scribe.
To understand how policy might respond, together with colleagues from Digital Rights Watch, Australian Community Managers, we convened a roundtable of academics, policy experts and practitioners across AI research, consumer law, online safety and women’s rights. The consensus was unanimous that determining the window for effective intervention is narrow, and the structure matters enormously.
The AI Plan contains several promising mechanisms: an AI Safety Institute, clever additions to Australian Consumer Law and the Online Safety Act, and investment in AI safety research all appear on the agenda. But each carries a significant risk if implemented without the people most affected by AI harms in mind.
An AI Safety Institute and more AI safety research are both essential, but without regulatory power, both would function as ‘talking shops’ rather than enforcement bodies. A better model would build on Australia’s reputation as a regional leader in digital regulation, giving the Institute real investigative authority and explicit community engagement functions.
Using consumer law to tackle AI risks also has its appeal, but plenty of limitations. AI products and services are rarely tested in the same way as traditional products, which makes it nearly impossible for regulators to establish that harm occurred because of a specific AI system. If companies can sell AI-embedded products to consumers without disclosing how those systems behave, standard consumer protection frameworks are largely unenforceable. The fix is requiring mandatory risk assessments, backed by strong investigative powers and genuine civil society capability to build an evidence base.
The imminent Digital Duty of Care—currently in development as a commitment to reviewing the Online Safety Act—requires digital service providers to evaluate their systems and processes’ effectiveness at preventing harm. But recently released plans outline that including AI in the Duty is not currently on the cards. AI products and services, including Generative AI and General Purpose AI will fall out of scope, unless they are a chatbot deployed by a social media platform.
And the Online Safety Act itself requires updating. Online gender-based violence and hate speech, when deployed at scale as information warfare during an election for example, already exceed the Act’s capacity to respond. Generative AI intensifies this. This is not individual harm. It is systematic. Neither the Online Safety Act nor our existing electoral integrity frameworks are currently equipped to address it.
Australia has demonstrated that digital regulation is achievable. Our track record creates both an opportunity and an obligation. The social media era has showed us what happens to women when government waits for harm to accumulate before acting. The evidence is already here. The question is whether the government will build something that works for everyone—or something that works for the industry that lobbied for it.
The experts are asking for action without delay. So are the women who are already experiencing harm.