AI image abuse puts women at risk as pressure grows for safer design

Content note: sexual and image-based abuse. Artificial intelligence is making it easier to fabricate intimate images and target people without their consent. Australian researchers and safety advocates say women and girls bear a disproportionate burden, and that responsibility rests with the people designing, deploying and regulating the technology rather than with an apparently independent machine.

The harm is not made unreal because a picture is synthetic. A fabricated image can humiliate someone, undermine their reputation, be used for threats or coercion, and remain accessible even when the victim and perpetrator are physically apart. The technology can extend the reach, persistence and scale of existing abuse.

What the research measures

A 2023 analysis by consumer digital-privacy platform Security Hero reported a 550 per cent increase in online deepfake videos between 2019 and 2023. It classified 98 per cent of the videos in its research as pornographic and said 99 per cent of those targeted women. Those are findings from that study and its dataset, not a new census of every AI-generated image or all internet content.

Australian National Research Organisation for Women's Safety research published in 2022 surveyed more than 4,500 adults. More than half reported at least one form of technology-facilitated abuse, rising to 71.5 per cent among 18-to-24-year-olds. One in four reported sexual or image-based abuse. Participants described fear, humiliation, depression and feeling controlled.

The four commonly reported categories included monitoring and controlling behaviour, emotional abuse, threats and harassment, and sexual or image-based abuse. The organisation's chief executive Tessa Boyd-Caine says abusers adapt quickly to tools such as location tracking, remotely controlled vehicles and everyday devices, so responses must keep pace with people's actual experiences.

The problem reaches public life as well. Social Services Minister Tanya Plibersek said in October that female politicians faced an unprecedented level of online violence. She said bots and inauthentic material contributed, with the effect of trying to silence women in politics.

Who designs AI, and whose voice is heard?

University of Melbourne law professor Jeannie Marie Paterson says female leaders are far more likely to be depicted in non-consensual synthetic intimate images. She says images can trigger a more immediate response than words because viewers tend to place trust in what they see, making fabricated visual material particularly damaging.

She also links the problem to unequal influence within the technology industry. Men remain prominent in development, company commentary and public discussion, while women are often concentrated in policy or marketing roles. eSafety Commissioner Julie Inman Grant says online abuse manifests differently for women and girls than for men.

A November 2025 UN Women report, co-authored by Everyday Sexism Project founder Laura Bates, found women made up only 30 per cent of AI professionals globally. Among 138 countries examined, 24 mentioned gender in national AI strategies and just 18 included substantive measures addressing effects on women and gender discrimination.

Human choices remain behind the systems

Paterson challenges descriptions of AI as simply learning on its own as it goes. Models are retrained at intervals and their operation reflects choices by developers and deployers. Describing the process as autonomous can obscure responsibility for how a system is built and what it is permitted to produce.

Developers can introduce safeguards and make different decisions about outputs. Paterson points to differences between systems such as Grok and Claude as evidence that design choices matter. Her position is that harmful material does not emerge in a policy vacuum beyond human influence.

She rejects the argument that preventing harm must be set aside to protect innovation. As with a vehicle, she argues, development should proceed within a duty not to injure people. In her view, responsibility and the values promoted by the technology matter as much as detailed rules.

Proposed rules and existing routes for help

People targeted by deepfake intimate images can seek help from police and the eSafety Commissioner. In certain circumstances the commissioner can require platforms to remove intimate material shared without consent. The existence of a removal route does not mean every case can be solved instantly or that every kind of harm falls within one power.

Inman Grant told an October parliamentary inquiry that AI services marketed as removing clothing from images were being downloaded about 200,000 times a month in Australia. She said her office had taken enforcement action against eight services. She wants stronger powers to test AI independently and assess safeguards, rather than relying on companies' own assurances.

The government's Digital Duty of Care legislation was introduced last month. The proposal would require online service providers to take reasonable steps to prevent harm and apply safety measures, with stronger enforcement, fines and content-removal powers. Those proposed requirements should not be confused with a statement that every provision is already law and operating.

Boyd-Caine says policy must address uses of synthetic media beyond pornography, including intimidation, humiliation and control. Paterson supports an active government response and wants companies to prevent harmful image creation at the design stage, including synthetic imagery used to facilitate abuse.

A public say in AI development

Paterson says ethics and public understanding should shape development. More women and other underrepresented groups need a say, while better AI literacy can help Australians understand systems and respond to their effects. She points to smaller models carefully tuned for useful tasks, such as helping people understand health or legal systems, as examples of potentially beneficial uses.

Her central point is that people retain the ability and the right to influence what these tools become. AI should benefit humans, and a public discussion about its risks should include the people most exposed to harm rather than leave the future entirely to technology companies.

For support relating to sexual assault or domestic and family violence, contact 1800RESPECT on 1800 737 732 or text 0458 737 732. In an immediate emergency in Australia, call 000.