MANILA, Philippines — Despite the scale of harm, women can’t get protection from artificial intelligence (AI) deepfake abuse because prosecutions are rare and platforms routinely fail to act, with survivors often re-traumatized when they try to seek help, according to the United Nations agency advancing women’s rights and gender equality.
Deepfakes are AI-manipulated images, audio or videos that make it appear someone said or did something they never did.
“The technology itself isn’t new. But its weaponization against women and girls is a newer phenomenon, and it’s accelerating fast,” UN Women said.
Deepfake sexual images now make up the vast majority of all deepfake content online, and almost all depict women.
The agency said the numbers are stark with 98 percent of all deepfake videos depicted women with non-consensual pornographic images, often referred to as deepfake nudes or fake nude images.
Deepfake videos were an estimated 550 percent more prevalent in 2023 than in 2019, and the tools to create them are widely available, usually free and require very little technical expertise.
Once posted, AI-generated content can be replicated endlessly, saved to private devices and shared across platforms, making it nearly impossible to fully remove.
UN Women explained that deepfake creators rarely face justice and the accountability gap is so wide and so gendered.
Less than half of countries have laws that address online abuse. Even fewer have legislation that specifically covers AI-generated deepfake content. Most “revenge porn” or image-based abuse laws were written before deepfakes existed, leaving gaping loopholes that perpetrators walk straight through.
The agency said that even when laws exist, enforcement frequently fails. Investigators need digital forensics expertise, cross-border coordination and platform cooperation to build a case, and most justice systems don’t have adequate resources for any of these.
“Evidence disappears fast as content spreads and copies multiply. Perpetrators hide behind anonymity or operate across jurisdictions. Platforms are slow – or unwilling – to share data with law enforcement, especially in cross-border cases. Digital forensics backlogs mean cases stall before they even get started,” the agency added.
Underreporting is one of the biggest barriers to accountability, it said.
“Many survivors choose instead to block, withdraw and try to survive, because taking down the content and protecting themselves is urgent. But that lets perpetrators off the hook,” it said.
Tech platforms, the agency said, have long hidden behind “intermediary” status to avoid responsibility for user-generated content.
In practice, this means platforms that are slow to remove abusive content, have opaque and inconsistent reporting processes, give automated rejection of takedown requests, and little to no cooperation with law enforcement.