Life-threatening information risks that can lead to serious harm
In Geneva on July 7, the United Nations refugee agency warned that artificial intelligence has transformed online hatred into a mechanism of physical harm, as deepfakes and algorithmically amplified misinformation increasingly precede violence against the world's 35.6 million refugees. Speaking at the AI for Good Global Summit, UNHCR advisor Gisella Lomax drew a direct line between fabricated digital content and real-world consequences — closed doors, broken communities, and in the gravest cases, killings and forced displacement. Her warning reflects a deeper tension of our age: the same technologies promising to democratize knowledge are being turned against those who have already lost the most. The agency's appeal to tech companies to treat refugee protection as an engineering priority worthy of genuine resources is, at its heart, a question about what obligations power carries.
- Generative AI has dramatically accelerated the spread of deepfakes and hate speech targeting refugees, outpacing the fact-checking systems designed to contain them.
- False narratives are not abstract harms — they have triggered protests, physical attacks, killings, and waves of forced displacement among populations already stripped of stability.
- Smugglers and traffickers exploit the same misinformation ecosystem, using digital platforms to lure desperate refugees into dangerous situations with false promises of safety.
- UNHCR is calling on tech companies to invest in content moderation tools that function in humanitarian contexts and in less-common languages currently ignored by major platforms.
- The agency frames this not as a debate about migration policy but as a life-threatening information crisis demanding urgent partnership between the humanitarian and technology sectors.
At the AI for Good Global Summit in Geneva on July 7, UNHCR senior advisor Gisella Lomax delivered a precise and troubling account of how online misinformation has evolved from reputational harm into physical danger. False narratives about refugees, she explained, close off access to jobs and education, erode the social trust needed for integration, and in extreme cases have directly preceded violence, killings, and forced displacement.
What distinguishes this moment is scale. Generative AI has introduced deepfake videos of UNHCR staff and refugees themselves — fabricated content that carries the appearance of evidence, spreads faster than corrections, and reaches communities where fact-checking infrastructure is nearly nonexistent. At the close of 2025, 117.8 million people were forcibly displaced worldwide, including 35.6 million refugees under UNHCR's mandate — two-thirds of them from Venezuela, Ukraine, Syria, Afghanistan, and Sudan. These are people already living without security, now targeted by information designed to cast them as threats or criminals.
Lomax was careful to separate legitimate debate about migration policy from what she described as life-threatening information risks. She also noted that the misinformation ecosystem enables predators: smugglers and traffickers use digital platforms to spread false promises, drawing vulnerable people into dangerous situations.
Yet her message was not one of pure alarm. AI, she argued, could also be a tool for managing humanitarian crises — if the companies that built these systems choose to deploy it that way. Her call was direct: tech platforms must partner with humanitarian organizations, invest in moderation tools that work in languages like Dari, Tigrinya, and Somali, and recognize the connection between a false post and a real death. Whether that appeal will be answered remains, for now, an open question.
In Geneva this week, the United Nations refugee agency delivered a stark warning: the lies spreading about refugees online are no longer just words. They are becoming weapons. Misinformation and hate speech, turbocharged by artificial intelligence, are inciting real-world violence against some of the world's most vulnerable people, and the tech companies that host these platforms have largely stood by.
The alarm was raised on July 7 at the AI for Good Global Summit, where Gisella Lomax, a senior advisor on information integrity at UNHCR, laid out the mechanics of harm with clinical precision. When false narratives take hold, she explained, they don't simply damage reputations. They close doors. They reduce access to jobs and education. They poison the social fabric that allows refugees to integrate into new communities. In the worst cases, they have preceded physical attacks, killings, and waves of forced displacement.
What makes this moment different—what makes it urgent—is the scale and speed that generative AI has introduced. Deepfake videos of UNHCR staff and refugees themselves are now circulating, presenting a challenge the agency did not face five years ago. These fabricated images and videos carry the weight of apparent evidence. They spread faster than corrections can catch them. They reach people in languages and regions where fact-checking infrastructure barely exists. The technology that was supposed to democratize information has instead become a tool for manufacturing consent for violence.
The numbers are staggering. At the end of 2025, there were 117.8 million forcibly displaced people worldwide. Of these, 35.6 million were refugees under UNHCR's mandate. Two-thirds of them came from just five countries: Venezuela, Ukraine, Syria, Afghanistan, and Sudan. These are people already stripped of home, security, and stability. They are now targets of a coordinated information assault that portrays them as threats, invaders, or criminals—often with no basis in fact.
Lomax was careful to distinguish between legitimate public debate about migration policy and what is actually happening. "This isn't about how societal concerns on migration and asylum are debated online," she said. "It's about life-threatening information risks." She cited the example of smugglers and traffickers who use digital platforms to spread false promises of safety and employment, luring desperate people into dangerous situations. The misinformation ecosystem doesn't just harm refugees directly; it enables the predators who exploit them.
Yet Lomax also offered a counterpoint that complicates the narrative of helplessness. Artificial intelligence, she suggested, need not be only a weapon. If deployed thoughtfully, it could help manage humanitarian crises. The question is whether the companies that have built these systems will choose to do so. Her call to action was direct: tech companies, AI developers, and digital platforms need to partner with humanitarian organizations, invest resources, and collaborate on solutions. Content moderation tools need to work in humanitarian contexts and in languages that most of the world's platforms currently ignore.
The gap between what is needed and what exists is vast. UNHCR is essentially asking the tech industry to treat refugee protection as a priority worthy of the same engineering resources devoted to engagement metrics and advertising algorithms. It is asking platforms to moderate content in Dari and Tigrinya and Somali with the same rigor they apply to English. It is asking companies to see the connection between a false post and a real death. Whether that ask will be answered remains an open question.
Citações Notáveis
The spread of misinformation, hate speech and deepfakes is exacerbating and inciting real-world harm to refugees and humanitarians.— Gisella Lomax, UNHCR senior advisor on information integrity
Our call to action is to all tech companies, AI companies and digital platforms to partner, invest and collaborate with humanitarian organisations.— Gisella Lomax, UNHCR
A Conversa do Hearth Outra perspectiva sobre a história
Why does misinformation about refugees matter more now than it did ten years ago?
Because the tools for spreading it have changed completely. A false rumor used to die in a village. Now it reaches millions in hours, and AI can manufacture evidence—deepfakes—that make the lie look true.
You mentioned that misinformation reduces access to jobs and education. How does that work?
When a community believes refugees are criminals or a drain on resources, they stop hiring them, stop welcoming their children in schools, stop renting to them. The lie becomes a barrier that's as real as a wall.
The source mentions that misinformation has triggered killings. Can you give a sense of how that chain works?
Online rumors create fear. Fear becomes anger. Anger becomes action. A false accusation spreads, a protest forms, and in the worst cases, it turns violent. The digital and physical worlds are not separate anymore.
Why would tech companies care about this? What's in it for them?
That's the hard part. Right now, there isn't much in it for them. Misinformation often drives engagement. Solving it requires resources and difficult choices. UNHCR is essentially asking them to prioritize human safety over profit.
Is there any reason to think they might listen?
Lomax said she sees "encouraging progress," though she didn't specify what that looks like. The fact that UNHCR is at this summit, making this case publicly, suggests they're trying to build pressure. But pressure and action are not the same thing.