On 15 May 2026, UN Women, UNESCO and the International Center for Journalists released Tipping Point: Online Violence Impacts, Manifestations and Redress in the AI Age, the second installment in the Tipping Point series and part of the UN Women ACT to End Violence Against Women programme. The findings are not subtle. Among the women journalists, activists and human rights defenders surveyed across 119 countries, 45% now self-censor in response to online violence — up from 30% in 2020. Roughly one in four report anxiety or depression linked to that abuse, and nearly 13% report symptoms consistent with PTSD. AI is named as an accelerant: 5% have been targeted with deepfakes or manipulated imagery, and a meaningful share have had intimate images shared without consent.
These are real harms with named victims, not a moral panic. The report's central diagnostic — that abuse designed to drive women out of public life is getting cheaper, faster and more convincing — deserves to be taken at face value. The harder question is what governments should do about it, and here the policy debate is at genuine risk of taking a wrong turn.
The redress gap is the actual problem
The most damning statistic is not about abuse volume but about impunity. UN Women's parallel finding, published in November 2025, is that 1.8 billion women and girls — 44% of the world's women — lack any legal protection from technology-facilitated abuse, and that fewer than 40% of countries have laws addressing cyber harassment or cyberstalking at all. Tipping Point echoes this: reporting to police roughly doubled between 2020 and 2025 (from 11% to 22%), and the share initiating legal action rose from 8% to 14%. Those increases sound like progress until you read them the other way — even now, roughly four in five incidents are never reported to police, and the overwhelming majority never reach a courtroom.
That is the failure worth fixing. When a deepnude of a journalist circulates and there is no statute under which she can compel its removal, no platform obligation to act on her notice, and no prosecutor with jurisdiction, the technology is not the binding constraint — the legal vacuum is.
What proportionate redress looks like
There is a workable template, and it is narrow by design. The United States' TAKE IT DOWN Act (S.146), signed into law on 19 May 2025, criminalizes non-consensual publication of intimate images — explicitly including AI-generated "digital forgeries" — and requires covered platforms to remove flagged material within 48 hours of notice, enforced by the Federal Trade Commission. Its compliance deadline for platforms fell on 19 May 2026, days after Tipping Point landed.
This is the right shape of intervention. It targets a defined, already-unlawful harm — non-consensual intimate imagery, real or synthetic. It gives victims a concrete remedy (a takedown right with a clock on it) rather than a vague platitude. And it pins liability on the parties — perpetrators and the platforms that ignore valid notices — instead of conscripting every user into a compliance regime. A law that says "remove this specific non-consensual image when the depicted person asks" is a scalpel. It is exactly the kind of demand-side accountability the report's redress gap calls for.
The wrong turn: treating abuse as a pretext for surveillance
The danger is that a report this stark gets used to justify interventions that do little for victims while quietly degrading the open internet for everyone. The strongest case for blunt instruments — mandatory age verification, identity requirements, default content scanning — is sincere: if you cannot identify who is online, the argument runs, you cannot deter or trace abusers, and children are exposed to the worst of it. That case deserves a fair hearing.
But the evidence does not support it. As EFF's Paige Collings documented on 5 June 2026, age-verification and identity mandates — now spreading across Australia, Indonesia, Malaysia, Brazil and the EU — force every user to hand over a face scan or passport to access ordinary services, building exactly the centralized identity honeypots that make doxxing and abuse worse. The UK's repeated failures show the verification technology does not reliably work. And these regimes do nothing to compel the 48-hour takedown of a specific deepfake. A woman journalist whose manipulated image is circulating is not helped by a law that ID-checks her readers; she is helped by a law that removes the image and reaches its publisher.
The distinction is not anti-regulation. It is the difference between regulation that closes the redress gap and regulation that expands state and platform surveillance under cover of protecting women. The report itself frames redress — legal remedies, platform accountability, mental-health and legal support — as the unmet need. It does not ask for the internet to be put behind an ID checkpoint.
The mandate is narrow accountability
The pro-innovation position is not that nothing should change. It is that the 1.8 billion-woman protection gap is best closed by laws that are specific, remedy-focused and liability-bearing — criminalizing non-consensual synthetic imagery, mandating prompt notice-and-removal, funding the cross-border cooperation that lets a journalist in one country pursue an abuser in another. Those measures are compatible with an open, low-friction, innovative internet. Identity mandates and broad scanning are not, and they do not even deliver the protection they promise. Tipping Point makes the harm undeniable. The task now is to answer it with precision rather than with the nearest available lever.