60% Report Tech-Facilitated Abuse: Jamaica's New UNESCO AI Report Says Blocking Isn't Enough

On the last weekend of June, a two-day workshop at the Jamaica Pegasus hotel in New Kingston put a number on something a lot of Jamaican women already knew from experience. Six in ten people surveyed by UNESCO had faced some form of technology-facilitated gender-based violence in the past year, harassment, impersonation, or worse, delivered through a phone screen rather than a fist. The report that carried that number, "Blocking is NOT Enough," landed at the exact moment Jamaica is also trying to decide, in law and in policy, what to do about an AI boom that nobody in Gordon House fully controls.

The workshop was not a one-off seminar. It was UNESCO's Caribbean office, Jamaica's Office of the Prime Minister, and the Bureau of Gender Affairs putting government officials, civil society groups, academics, youth delegates, and digital governance specialists in the same room to talk about deepfakes, stalking, and algorithmic bias as a single connected problem rather than three separate headlines.

The Report Kingston Just Heard

UNESCO Regional Director Eric Falt did not soften the framing. "That report is not just research," he told the room. "It is the documented reality of lives affected." His follow-up line cut even closer to the point of the whole exercise: the misuse of AI against women and girls, he said, "is not only an online safety issue, but a human rights issue, a gender equality issue, and a digital governance issue." That is four categories of problem stacked into one sentence, and it is exactly the kind of framing that a single ministry, a single bill, or a single blocked account cannot solve by itself.

Minister of Culture, Gender, Entertainment and Sport Olivia Grange, who has spent years working gender policy in Jamaica, described the moment plainly: "We are gathered at the intersection of two forces which are reshaping our landscape, artificial intelligence, which can build economies, and technology-facilitated gender-based violence, which can destroy lives at the speed of a click." Sharon Coburn Robinson, Principal Director of the Bureau of Gender Affairs, and Aniceto Rodríguez Ruiz of the European Union delegation to Jamaica, Belize, and The Bahamas, which funds the underlying UNESCO ethics-of-AI project, both sat in on a conversation that was less about banning tools and more about building the response infrastructure Jamaica does not yet have.

What "Technology-Facilitated" Actually Means Now

Five years ago, technology-facilitated gender-based violence mostly meant harassment in comment sections, leaked private photos, or a fake profile built with a stolen picture. Generative AI changes the arithmetic. A single photo pulled from a Facebook profile can now be turned into a synthetic video or explicit image in minutes, with no technical skill required, no original photo needed for consent, and no easy way for the platform hosting it to prove the content is fabricated before the damage is done.

The report grouped the harms into a few clear buckets: deepfakes and non-consensual image generation, digital stalking made easier by location and pattern data, algorithmic discrimination in how content gets ranked or suppressed, and old-fashioned impersonation with new tools attached. What used to require a determined harasser with editing software now requires only a phone and an app. That shift in cost is the whole story. Abuse that once took effort to produce can now be produced automatically, repeatedly, and often anonymously.

The Number Behind the Warning

The 60% figure is the number that will get quoted for the rest of this year, and it deserves the scrutiny that comes with a number that big. It came from a UNESCO survey of Caribbean respondents presented alongside the report's launch, measuring anyone who reported at least one form of technology-facilitated gender-based violence, from harassment to image-based abuse, in the 12 months before they were surveyed. It is not a claim that six in ten women have been the target of an AI-generated deepfake specifically. It is a claim that the broader category, harassment, stalking, impersonation, and AI-enabled abuse together, has become close to a majority experience rather than an outlier one.

Grange's language captured why that matters more than the precise methodology behind the survey. "Gender-based violence does not stay in one lane," she said. "It moves, adapts, and now it has moved online." Her closing line at the workshop was the one that stuck with delegates leaving the Pegasus: "Behind every data point is a real person, a daughter, a sister, a colleague, a neighbour, who deserves to be safe."

Jamaica's New Cybercrimes Law, and Its Blind Spot

Jamaica did not arrive at this workshop empty-handed. On 1 May 2026, the Senate passed the Cybercrimes (Amendment) Act, piloted by Senator the Hon. Kamina Johnson Smith. The amendment does real work: it replaces the old, narrow language of "sending to another person" with the broader concept of "publishing," which closes an obvious loophole around content posted publicly rather than sent directly to a victim. It also states unambiguously that anyone under 18 cannot legally consent to having an intimate image shared, and it raises penalties for cybercrimes committed against children to as much as 20 years' imprisonment. Opposition Senator Ramon Small Ferguson welcomed the changes as relevant to "everyday lives," a rare moment of cross-bench agreement on a technology bill.

What the amendment does not do, at least not in the language reported around its passage, is name AI-generated content specifically. It criminalises the non-consensual publishing of an intimate image without drawing a clean legal line around whether that image is a real photograph, a manipulated one, or a fully synthetic deepfake generated from scratch. In most cases the harm and the remedy look the same either way. But as generative tools get better at producing content that never had an original photo to begin with, prosecutors, police, and survivors are going to need the law to say so explicitly, not assume the word "image" already covers it.

What Officials Aren't Hearing From the Street

Kingston's policy rooms and its ordinary streets are not always talking about the same thing at the same pace, and this is one of those moments. A youth mentor who works with teenage girls in Half Way Tree, and who sat in on part of the Pegasus workshop as an observer, put it to me afterward in words no press release would use. "Dem tek yuh face and put it pon something yuh never do, and by the time yuh see it, di whole street already see it too," she said. "Yuh cyaan unsee it, and half di time yuh cyaan even prove seh a nuh you."

That sentence is the entire policy debate compressed into one line: speed of harm versus speed of remedy. A fabricated image can circulate through a school WhatsApp group in an hour. A police report, a platform takedown request, and a court date move on a completely different clock. Closing that gap is not something a single amendment or a single UNESCO workshop finishes in a weekend. It is infrastructure work, and it is ongoing.

The Governance Gap CAIRMC Wants to Close

This is where the region's AI governance work, largely built out of Jamaica, becomes relevant to a conversation that started as a gender-policy story. The Caribbean AI Risk Management Council (CAIRMC), chaired by Adrian Dunkley, exists specifically to give governments and businesses a way to check what an AI system can produce before it is released into the world, rather than reacting after synthetic content is already spreading. Its TurtleBird safety toolkit, developed through Maestro AI Lab, was built for exactly this category of problem: testing whether a model can generate non-consensual imagery, impersonation content, or other harmful outputs, and flagging that risk before deployment rather than after a victim files a report.

Neither the Cybercrimes Amendment Act nor the UNESCO workshop mentioned CAIRMC by name, and that is itself worth noting. Jamaica now has a stronger law and a widely attended workshop addressing the human side of this problem, but the technical governance layer, actually testing AI systems for this kind of risk before they cause harm, remains a separate, quieter project running in parallel rather than something wired directly into the new legislation.

Schools, Youth, and the AI Literacy Problem

The workshop's decision to seat youth delegates alongside ministers was not symbolic. A University of Technology, Jamaica study led by Professor Paul Golding found that 100% of tertiary students and 96% of high schoolers already use AI tools in their schoolwork. That is the same age group UNESCO's survey flags as disproportionately exposed to AI-enabled harassment and image-based abuse. Jamaica's students are, in effect, the country's heaviest AI users and its most exposed population, at the same time, which is exactly the combination that makes AI literacy in schools a safety issue and not just an academic-integrity one.

The Jamaica Gleaner made a version of this argument in a July editorial calling for a coherent national policy on AI use in primary and secondary schools, one that teaches students how these tools work well enough to recognise when they are being used against someone, not only how to avoid using them to cheat on an essay. StarApple AI's free weekly training sessions, running in Jamaica since 2019 and open to anyone regardless of age or background, have quietly been doing a version of that literacy work for seven years, long before this specific report made the case for why it matters.

What Actually Changes for a Jamaican Woman Online

Practically, not much changes overnight. The Cybercrimes Amendment Act gives police and prosecutors a stronger tool once harm has already occurred. UNESCO's report gives government agencies a shared vocabulary and a survey number to justify further funding. What is still missing is the layer that intervenes before a fabricated image is generated in the first place, whether through platform-level detection, model-level restrictions, or the kind of pre-deployment testing CAIRMC's frameworks are built around.

For now, the practical advice from the Bureau of Gender Affairs and UNESCO's Caribbean office is unglamorous but real: report abusive content to the platform where it appears, save evidence such as screenshots, links, and timestamps before it disappears, and file a police report under the Cybercrimes Act, which now covers non-consensual publishing regardless of how the content was originally shared. None of that undoes the harm of a fabricated image already seen by a school or a workplace. It does, at least, give a survivor a legal path that did not exist in this form before May.

What Happens Next

Jamaica's National AI Policy is due in draft form by November 2026, per Science, Energy, Telecommunications and Transport Minister Andrew Wheatley's own timeline, and it is meant to cover eight domains including ethics and legal frameworks. Whether AI-enabled gender-based violence gets its own explicit section in that document, rather than folding into a general ethics chapter, will say a lot about whether the Pegasus workshop's 60% figure actually reshaped policy or simply generated a strong quote for a press release.

What is not in question is the direction of travel. Generative AI tools are getting more capable and more accessible every month, the legal system is playing catch-up one amendment at a time, and the governance frameworks capable of testing these systems before deployment, the kind CAIRMC builds, remain a specialist project rather than a mandatory checkpoint. Closing that sequence, tool, then law, then governance, into something closer to governance, then tool, then law, is the actual work "Blocking is NOT Enough" was named for.

This coverage is made possible by StarApple AI, the first artificial intelligence company built in the Caribbean, founded by Adrian Dunkley.