The manosphere’s worship of synthetic intimate image abuse
The crowd behind the image: what 4chan tells us about synthetic abuse, and why Australia can’t look away
We tend to imagine image-based abuse as a private crime: one perpetrator, one victim, a secret kept until it isn’t. The picture reassures us, because it casts the harm as an aberration, the work of a single person who went wrong somewhere. What it obscures is how much of this abuse is produced in company, solicited openly, made to order, and applauded by an audience that treats the result as a contribution. The injury that reaches my counselling room as something intensely private often began as a collective performance.
I’m a counsellor and social worker, and much of what comes through my door is the slow aftermath of someone losing their sense of safety in their own body and their own image. Most people describe it the way the rest of us do, as a “deepfake,” a word that sounds like a problem for engineers. It is anything but abstract to the person it happens to. It reshapes how they dress, who they trust, whether they will stand in front of a camera at all. And the technical framing quietly works in the perpetrators’ favour, holding our attention on the tool and drawing it away from the people who choose to wield it and the audience urging them on.
That audience is what the Institute for Strategic Dialogue set out to document. When it published its analysis of how synthetic non-consensual intimate imagery is discussed and organised on 4chan, I didn’t read it as a policy paper. I read it as an account of the conditions that produce both the people I sit across from and the people who harm them. Its real contribution isn’t what it reveals about the technology. It’s what it reveals about the company that technology keeps, and about how ordinary cruelty becomes social currency once enough people are watching.
What the research found
The Dispatch, written by Leonie Oehmig and published in May 2026, looks at how the creation of synthetic NCII is talked about on 4chan. It draws on a qualitative reading of posts, backed by keyword analysis of 7,616 posts gathered over roughly three months. The scope is narrower than the subject, and the authors are careful about that, so I’ll be careful too. This is a study of how people talk, not a count of how many images were made. A keyword approach can’t fully rule out false positives, and because the analysis is text-based rather than image-based, it isn’t always possible to tell whether a post refers to a real person or a fictional character. The prevalence figures carry those caveats. What the posts show about how this abuse is requested, normalised and excused doesn’t depend on the counting, because it sits in the language itself.
Three patterns come out of it.
The first is how the requests are phrased. The researchers found that 2,927 posts, around 38 percent of the dataset, contained at least one keyword tied to directly soliciting abusive content: commands like “give her,” “make her,” “nudify,” usually attached to an uploaded photo of a real person and followed by degrading instructions. What gets me is how routine it is. Nobody in these threads is arguing about whether the thing is acceptable. The grammar already assumes it is. You don’t issue something as a flat instruction unless you’ve stopped seeing a person on the other end of it.
The second pattern is what the community does with the people who make the images. It celebrates them. Creators get called “wizards” or “wiz,” a term that turns up 777 times, which frames producing this material as a skill worth admiring. Requesters call them “Boss” or “Master,” ask them to “do your magic,” offer “praise and adoration” in exchange. Running through all of it is a chummy language of brothers and friends, a sense of belonging built on top of the abuse. This is the part I’d want people to sit with longest. The harm isn’t a flaw in the culture of these spaces. It’s the thing the culture is organised around. Making it earns you status, which is how you know it’s been normalised rather than merely tolerated.
The third pattern is how people justify it. The researchers found a recurring set of stories told openly on the platform, most of them some version of victim-blaming: a woman who put images of herself online has, supposedly, already consented to whatever’s done to them. Some posts cheer the technology specifically for stripping women of control over their own image. Others dress the abuse up as the natural order of things, women’s bodies framed as available to men by default, or as something the target had coming. And there’s the language of revenge, which the researchers rightly call misleading, because the word implies the person did something to earn it when nothing earns it.
If you’ve worked in this field, none of those excuses will be new. They’re the same scripts that run through family violence, sexual assault, coercive control. The tools are new. The thinking underneath them is old and familiar.
Why this is a coercive control story
That last point is where my own work and this research line up most closely.
We’ve spent years learning to see coercive control as a pattern rather than a single incident: the monitoring, the isolation, the slow rewriting of someone’s reality until they doubt what they saw with their own eyes, the treatment of a partner as property whose image and body and movements are yours to direct. Synthetic NCII is that same logic, only externalised and sped up. The researchers describe the technology as an extension of existing power relations that lowers the barrier to producing abusive material while widening its reach, and that’s exactly how it reads from where I sit. What once needed physical access to a person, or at least a private image of them, now needs a public photo and a few minutes.
The research also names something I see constantly in coercive control, which is abuse aimed at a person through the people they love. The researchers describe synthetic NCII used as retaliation, pointed not only at targets but at third parties because of their closeness to them, with women treated as collateral. One example was a request to “nudify” a woman and her daughter as a way of getting at the husband. Anyone who’s supported a person leaving a controlling relationship will recognise that move on sight. The woman was never really the target. She was the instrument, and her humiliation was a message meant for someone else. That’s part of what makes this so hard to capture in laws built around a single victim and a single image.
I work within a trauma-informed and queer-affirming frame, and there’s a limit in the research that I want to name, partly because the authors name it themselves and partly because it matters for the people I see. The dataset was overwhelmingly built around binary gendered language, and a keyword search for terms targeting gender-diverse or non-binary people returned only 29 posts, none of them used in the relevant context. The authors are honest that this is a limitation, not evidence of absence, and from clinical experience I’d put it more strongly. That a particular community’s vocabulary is binary tells you about that community’s vocabulary, not about who actually gets hurt. Trans and non-binary people, and queer people more broadly, are frequent targets of image-based abuse. Their near-absence from this dataset is a gap in the lens, not in the harm, and whoever takes this research further should go looking on purpose.
Where the regulation breaks down
The most practically useful part of the Dispatch is its argument about platform accountability.
The core problem the researchers identify is that the EU’s Digital Services Act sorts platforms mainly by size rather than by risk. Because 4chan sits below the threshold to be designated a Very Large Online Platform, it carries lighter obligations, mostly acting on notices about illegal content rather than the proactive risk assessment expected of bigger services. As the researchers point out, on an anonymous, lightly moderated platform, leaning on users to flag content is never going to surface much at scale. You can’t build a safety regime around the hope that people will report the very thing the space exists to do.
Underneath that sits a more basic failure. The researchers note that 4chan appears not to have appointed a legal representative within the EU, which Article 13 of the DSA requires of providers offering services there, and that because of this it’s unclear which national regulator should even take the lead. The accountability gap has gone unaddressed. There’s a parallel process in the UK: the Dispatch notes that Ofcom has taken enforcement action over 4chan’s failure to meet its obligations under the Online Safety Act, including a failure to respond to a statutory information request and a lack of age assurance, with the outcome still pending when the piece was written.
And the harm doesn’t stay in one place. The researchers found 4chan working as an entry point, with conversations moving off to more private spaces. Across the dataset they counted references to Kik, Teleguard, Discord and Telegram, and noted they were only catching explicit mentions, so the real number is probably higher. This is the migration problem, and it’s the part regulators should worry about most. You can clean up the shopfront and watch the trade move to the back room.
Where Australia actually sits
The ISD research is focused on Europe and the UK, but I’m reading it from Victoria, so the obvious question for me is what it means here. The answer is more encouraging than you might expect on the law, and more sobering on everything underneath it.
Australia has moved fairly decisively. At the Commonwealth level, the Criminal Code Amendment (Deepfake Sexual Material) Act 2024 strengthened the offences targeting non-consensual sharing of sexually explicit material online, including AI-generated and altered imagery, with a maximum penalty of six years, and seven for the aggravated offence of creating or altering the material. That law has already been tested. In 2026 it produced its first guilty plea, in a case involving intimate images transmitted on X without consent.
My own state is, on paper, ahead of much of the country. Victoria introduced offences in 2022, through the Justice Legislation Amendment (Sexual Offences and Other Matters) Act, covering producing, distributing and threatening to distribute an intimate image contrary to community standards, with production punishable by three years, and the offence explicitly reaches digitally superimposing someone’s face onto a naked body. That’s worth saying plainly, because it answers one of the gaps the ISD research keeps circling. In Victoria the creation of this material, not just the sharing of it, is already a crime. A 2025 New South Wales parliamentary analysis, by contrast, found that the non-consensual creation of sexually explicit deepfakes of adults remained unprohibited there, with the most relevant offences dating from 2017 and aimed at recording and distribution. Protection in this country is patchy, and which side of a state border you live on still changes what the law will do for you.
Across all of it sits the eSafety Commissioner, which gives Australia something the EU and UK frameworks lack: a civil scheme aimed squarely at getting content taken down. The Commissioner treats deepfake pornography against adults as image-based abuse, runs a complaints scheme covering both real and synthetic material, and reports a high success rate on takedowns. The Commissioner has also been blunt about the scale and the gendering of it, pointing out that pornographic material makes up the overwhelming majority of deepfake content online and that almost all of it depicts women and girls.
So the Australian picture really is better than the regulatory vacuum the ISD piece describes. I’d still resist taking too much comfort from it, for two reasons that come straight out of the research.
The first is that the migration problem doesn’t respect our borders any more than Europe’s. A takedown scheme is reactive by design. It works on content that’s already surfaced somewhere visible. The whole point of the ISD findings is that the coordination happens upstream, in places like 4chan, and then slips into closed channels where neither the eSafety Commissioner nor anyone else can easily see it. Australian law can reach the Australian who shares an image. It can do very little about the anonymous post on a foreign imageboard that started it, or the encrypted group it moved into.
The second is the one I care about most. The law is the floor, not the ceiling. The ISD research is, at bottom, an argument that this abuse is socially produced. It’s incentivised, rewarded and excused by a community that has built a whole status hierarchy around it. You can’t prosecute your way out of a culture. You can punish individual acts after they happen, but the thing generating those acts is a set of beliefs about what women are for, and those beliefs aren’t illegal. They sit upstream of the law, and that’s where the real work is.
What I’m left with
I keep coming back to the wizards. Seven hundred and seventy-seven times, someone reached for a word that turns making a humiliating image into a kind of magic and the person who made it into someone admirable. That’s not a technical detail, it’s a moral one. It tells you that for the people in these spaces, the cruelty isn’t a cost they’re grudgingly paying. It’s the product. It’s the thing that earns the praise.
For those of us who work with the people on the receiving end, the research is valuable precisely because it refuses to treat synthetic NCII as a story about clever software. It’s a story about contempt, organised and rewarded at scale, using tools that happen to be new. The victim-blaming, the entitlement, the use of a person as a message to send to someone else: I’ve heard all of it in a counselling room, long before anyone could generate an image in thirty seconds. The technology lowered the barrier. It didn’t invent the wanting. Until our responses, legal and social and clinical alike, take the wanting as seriously as the tool, we’ll keep meeting the same harm wearing a slightly newer face.
If you’ve been affected by image-based abuse in Australia, the eSafety Commissioner runs a reporting and removal scheme at esafety.gov.au, and 1800RESPECT (1800 737 732) offers confidential support for sexual, domestic and family violence.
Sources: Institute for Strategic Dialogue, “High risk, small platforms: Understanding synthetic non-consensual intimate imagery and community norms on 4chan,” by Leonie Oehmig, Digital Dispatch, 21 May 2026. Australian legal and regulatory detail drawn from the Criminal Code Amendment (Deepfake Sexual Material) Act 2024 (Cth), the Justice Legislation Amendment (Sexual Offences and Other Matters) Act 2022 (Vic), the NSW Parliamentary Research Service (Research Paper 2, 2025), and the Office of the eSafety Commissioner.