Life-threatening information risks that can lead to serious harm
At a Geneva summit meant to explore technology's humanitarian promise, the United Nations refugee agency issued a warning that cuts to the heart of the digital age's deepest contradiction: the same tools built to connect humanity are being turned against its most vulnerable members. UNHCR reports that AI-accelerated misinformation and deepfakes are not merely shaping opinion about the world's 117.8 million displaced people — they are triggering protests, violence, and forced displacement in communities that have already survived one uprooting. The agency's appeal to tech companies is, at its core, a question about moral responsibility: when a platform profits from the architecture of attention, who bears the cost when that architecture is weaponized against the powerless?
- Generative AI is producing deepfakes of refugees and aid workers at a pace and convincingness that outstrips any existing humanitarian response.
- False narratives are closing real doors — refugees are losing access to jobs, schools, and community belonging as coordinated disinformation poisons host societies against them.
- The violence is documented and escalating: misinformation has been a direct factor in protests turning deadly, physical attacks, and people being forced from homes they had only just found.
- Smugglers and traffickers have seized on digital platforms to deceive desperate people with false promises, funneling them toward exploitation rather than safety.
- UNHCR is calling on tech companies to invest in content moderation that works in humanitarian emergencies and in the less-common languages where the most vulnerable people communicate.
- Some collaborative efforts between platforms and humanitarian organizations exist, but they remain scattered and far too small for a crisis affecting nearly 118 million people.
The United Nations refugee agency issued a sharp warning this week: artificial intelligence is accelerating the spread of misinformation and hate speech in ways that place refugees in direct physical danger. The alert came as UNHCR participated in the AI for Good Global Summit in Geneva, a gathering designed to explore how technology might address rather than deepen global crises.
Gisella Lomax, UNHCR's senior adviser on information integrity, argued that major displacement crises are almost always paired with information crises. Deepfake videos, false narratives, and coordinated disinformation campaigns are not merely political noise — they move offline. Protests turn violent. Attacks happen. In documented cases, misinformation has directly contributed to people being forced from their homes a second time.
The numbers frame the stakes. At the end of 2025, 117.8 million people were forcibly displaced worldwide, with 35.6 million classified as refugees under UNHCR's mandate. Two-thirds came from just five countries: Venezuela, Ukraine, Syria, Afghanistan, and Sudan. Each population faces not only the original displacement but a second wave of false information engineered to turn host communities against them.
Generative AI has sharpened this threat. Deepfakes of UNHCR staff and refugees are growing more convincing. Smugglers now use digital platforms to lure vulnerable people with false promises of safety and work, delivering them instead into exploitation. At scale and speed, distorted information narrows access to jobs, closes educational doors, and fractures the social trust that integration requires.
Lomax drew a careful line between legitimate debate about migration policy and what she called life-threatening information risks — the dehumanizing rhetoric and coordinated campaigns designed to convert public opinion into mob action. Trusted information, she insisted, is not a luxury in humanitarian settings. It is a survival requirement.
UNHCR's ask is concrete: tech companies and AI developers must partner with humanitarian organizations, build content moderation tools that function in emergency contexts and in languages beyond English and Mandarin, and scale up what fragmented progress already exists. The agency's position is not that AI is inherently harmful — handled well, it could strengthen humanitarian response. But without deliberate intervention, the current path leads toward more deepfakes, more coordinated disinformation, and more real-world violence against people who have already lost nearly everything.
The United Nations refugee agency delivered a stark warning this week: artificial intelligence is turbocharging the spread of misinformation and hate speech in ways that are putting refugees in direct physical danger. The alert came as UNHCR participated in the AI for Good Global Summit in Geneva, a gathering meant to examine how the technology might solve rather than compound global crises.
The problem, according to Gisella Lomax, UNHCR's senior adviser on information integrity, is that major displacement crises are almost always accompanied by what she calls "information crises." False narratives, deepfake videos, and coordinated disinformation campaigns are not merely annoying or politically charged—they are inciting real-world harm. When rumors spread online about refugees, when dehumanizing speech takes hold, when scapegoating becomes systematic, the consequences move offline. Protests turn violent. Attacks occur. In documented cases, misinformation has been a factor in forcing people from their homes.
The scale of the problem is immense. At the end of 2025, there were 117.8 million forcibly displaced people worldwide. Of those, 35.6 million were refugees under UNHCR's mandate, with two-thirds coming from just five countries: Venezuela, Ukraine, Syria, Afghanistan, and Sudan. Each of these populations faces a dual crisis—the displacement itself, and then a second wave of false information designed to turn host communities against them or to exploit their vulnerability.
Generative AI has made this worse. Deepfakes of UNHCR staff and refugees are becoming more common and more convincing. Smugglers and traffickers now use digital platforms to spread false promises of safety and employment, luring desperate people into situations of exploitation and danger. When information is distorted at scale and at speed, the collateral damage extends into every corner of refugee life. Access to jobs narrows. Educational opportunities disappear. Integration becomes harder. Social cohesion fractures.
Lomax was careful to distinguish between legitimate public debate about migration policy and what she called "life-threatening information risks." The latter category includes the false accusations, the dehumanizing rhetoric, the coordinated campaigns designed to turn public opinion into mob action. She emphasized that trusted information is not a luxury in humanitarian contexts—it is a necessity for survival.
The call from UNHCR is direct: tech companies, AI developers, and digital platforms need to partner with humanitarian organizations. They need to invest in content moderation tools that actually work in humanitarian emergencies and in languages that are not English or Mandarin. They need to systematize and scale up the progress that has already begun. Lomax acknowledged that some encouraging work is underway, but it remains fragmented and insufficient.
The agency's position is that AI itself is not the enemy. Handled correctly, the technology could help manage humanitarian crises more effectively. But without intervention—without deliberate choices by the companies that build and profit from these platforms—the current trajectory is toward more deepfakes, more coordinated disinformation, more real-world violence against people who are already among the world's most vulnerable.
Notable Quotes
This isn't about how legitimate societal concerns on migration and asylum are debated online—but about life-threatening information risks that can lead to serious harm in humanitarian contexts.— Gisella Lomax, UNHCR senior adviser on information integrity
Our call to action is to all tech companies, AI companies and digital platforms to partner, invest and collaborate with humanitarian organizations.— Gisella Lomax, UNHCR
The Hearth Conversation Another angle on the story
When you say misinformation is "life-threatening," are you talking about the psychological toll, or something more direct?
Something more direct. False rumors have triggered mob violence. Dehumanizing speech has preceded physical attacks. In extreme cases, misinformation has been documented as a factor in forced displacement—people fleeing not just conflict but the communities that turned against them based on lies.
How does AI change the equation compared to, say, five years ago?
Speed and scale. A human can spread a rumor. An AI can generate a convincing deepfake video of a refugee committing a crime they didn't commit, and distribute it to millions in hours. It's not a new problem; it's an old problem on steroids.
The UN is asking tech companies to help. Why would they?
That's the real question. Right now there's no financial incentive. Engagement drives ad revenue, and outrage drives engagement. UNHCR is essentially asking platforms to voluntarily reduce the content that makes them money.
What about content moderation tools that already exist?
They exist, but mostly for major languages and wealthy markets. A deepfake in Arabic or Tigrinya or Dari might never get flagged because the moderation systems weren't built for those languages. That's where the gap is.
So this is partly a resource problem?
It's partly that. But it's also a priority problem. Refugees aren't a profitable demographic for tech companies. They're not the target market. So the tools don't get built for them.