Information so distorted it creates life-threatening risks
In Geneva, the United Nations has placed before the world's most powerful technology companies a moral reckoning: the same artificial intelligence reshaping human possibility is also accelerating the dehumanization of 117.8 million displaced people, turning online falsehoods into physical violence, displacement, and death. The UN refugee agency's warning at the AI for Good Global Summit is not merely a policy appeal — it is a reminder that information, like water and shelter, has become a condition of survival for the world's most vulnerable. History has long shown that dehumanizing speech precedes atrocity; what is new is the speed and scale at which AI now carries that speech across borders and languages.
- Deepfake videos, fabricated narratives, and AI-amplified hate speech are no longer abstract threats — they have already triggered protests, attacks, killings, and the forced displacement of refugees who survived one crisis only to face another online.
- With 117.8 million people forcibly displaced globally, the populations most targeted — Venezuelans, Ukrainians, Syrians, Afghans, Sudanese — are also the least protected by content moderation systems built for dominant languages and stable societies.
- Smugglers and traffickers are exploiting the same digital chaos, using deceptive platforms to lure desperate people with false promises of safety, turning the information crisis into a direct pipeline for exploitation.
- The UN is drawing a deliberate line between legitimate debate about migration policy and weaponized misinformation that creates life-threatening conditions — insisting this is a humanitarian protection issue, not a free speech debate.
- Progress exists but remains fragmented and unscaled; UNHCR is calling on tech giants to treat information integrity as essential humanitarian infrastructure, investing in moderation tools built for crisis contexts and less-common languages.
At the AI for Good Global Summit in Geneva, the United Nations issued an urgent warning: artificial intelligence is dramatically accelerating the spread of misinformation and hate speech targeting the world's displaced populations, and the consequences are no longer virtual.
By the end of 2025, 117.8 million people were forcibly displaced globally — 35.6 million of them refugees under UN protection, two-thirds originating from Venezuela, Ukraine, Syria, Afghanistan, and Sudan. These are the communities most exposed to what UN refugee agency advisor Gisella Lomax described as a weaponized information environment. Online rumors and dehumanizing narratives have incited protests and attacks, preceded physical violence and killings, and in documented cases contributed directly to further displacement. Generative AI has added a new dimension: deepfake videos of refugees and humanitarian workers can now be produced at scale, a threat that barely existed a few years ago.
The harm extends beyond violence. Misinformation closes doors — to employment, education, and community integration. Smugglers exploit the disorder, using digital platforms to spread false promises of safety to desperate people, funneling them into dangerous situations. In active conflict zones, these dynamics move at lethal speed.
Lomax was careful to separate this concern from debates about migration policy, which she acknowledged are legitimate and ongoing. The UN's focus is narrower and more urgent: speech so distorted and dehumanizing that it creates direct physical risk — a pattern history has repeatedly shown to precede mass violence.
The UN's call to action is clear: tech companies and AI developers must partner with humanitarian organizations, build content moderation systems that function in crisis settings, and invest in tools for the less-common languages where many displaced communities communicate and where misinformation spreads most freely. Some cooperation is already visible, Lomax noted, but it remains scattered rather than systemic. The summit represents a broader UN effort to ensure AI serves human protection rather than undermines it — but as the agency made plain, potential without coordinated action changes nothing for the people already at risk.
At the AI for Good Global Summit in Geneva this week, the United Nations delivered a stark warning: artificial intelligence is turbocharging the spread of false information and hate speech that directly harms refugees, and the world's tech companies need to act now to stop it.
The scale of the problem is enormous. At the end of 2025, there were 117.8 million people forcibly displaced worldwide—68.7 million trapped within their own countries, and 35.6 million refugees under UN protection. Two-thirds of those refugees come from just five places: Venezuela, Ukraine, Syria, Afghanistan, and Sudan. These are the populations most vulnerable to the information wars playing out online.
Gisella Lomax, the UN refugee agency's senior advisor on information integrity, laid out the mechanics of the harm during a briefing with reporters. When false narratives and dehumanizing speech spread unchecked, they don't stay abstract. Online rumors and scapegoating have triggered real protests and attacks. In documented cases, they have preceded physical violence and killings. Misinformation has even been a factor in forcing people from their homes. The problem is compounded by generative AI, which can now produce deepfake videos of refugees and humanitarian workers at scale—a growing challenge that didn't exist just years ago.
The consequences ripple through every part of a refugee's life. When misinformation takes hold, it closes doors. People lose access to jobs and education. Integration into new communities becomes harder. Social cohesion fractures. Smugglers and traffickers exploit the chaos, using digital platforms to spread false promises of safety and employment, luring desperate people into dangerous situations. In conflict zones and humanitarian emergencies, these information risks move at lethal speed.
Lomax was careful to distinguish between legitimate public debate about migration and asylum policy—which happens online constantly—and what the UN is flagging here: information so distorted and weaponized that it creates life-threatening risks in humanitarian contexts. The distinction matters. This isn't about free speech. It's about false accusations and dehumanizing speech that have historically preceded violence.
The UN's call is direct: tech companies, AI developers, and digital platforms need to partner with humanitarian organizations, invest resources, and collaborate to turn this tide. Part of that means building content moderation tools that actually work in humanitarian crises and in languages that aren't English or Mandarin—the less-common languages where many displaced populations communicate and where misinformation often spreads unchecked.
Lomax acknowledged that some progress is visible. But it's scattered, not systematic, not scaled. What's needed is a coordinated effort that treats information integrity as essential infrastructure for refugee protection, the way the UN treats water and shelter. The summit itself is part of a broader UN push to harness AI's potential to solve global problems rather than amplify them. But potential means nothing without action. The question now is whether the tech giants will answer the call.
Citações Notáveis
Online rumors, false accusations, scapegoating and dehumanizing speech have triggered protests and attacks. In extreme cases, they have been linked to physical violence, killings, and are even a factor in forced displacement.— 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 the UN think AI specifically is making this worse? Couldn't misinformation spread just as easily without it?
It could spread, yes, but not at this speed or scale. Deepfakes used to require expensive equipment and expertise. Now generative AI can produce them in minutes. A false video of a refugee or a humanitarian worker can reach millions before anyone can fact-check it. That's the acceleration.
You mentioned that misinformation has led to killings. Can you walk through how that actually happens?
It starts with a false narrative—say, blaming refugees for a crime they didn't commit. That spreads online, people believe it, protests form, tensions rise. In some cases, mobs have attacked refugee settlements or individuals based on rumors that turned out to be completely false. The information creates the conditions for violence.
The UN is asking tech companies to help. Why would they? What's their incentive?
That's the hard question. Right now, there's no strong incentive. Engagement drives their business model, and inflammatory content drives engagement. But there's also reputational risk, legal risk in some jurisdictions, and moral pressure. The UN is essentially saying: you have the tools to moderate this, you have the responsibility to use them.
What about the languages issue? Why is that important?
Most content moderation happens in English and a few other major languages. But refugees often communicate in smaller languages—Tigrinya, Dari, Rohingya dialects. Misinformation spreads in those spaces with almost no oversight because the platforms don't have moderators who speak those languages. It's a blind spot.
Is there any sign this is actually changing?
Lomax said there's encouraging progress, but it's not systematic yet. Some platforms are starting to work with humanitarian groups. But it needs to be scaled up dramatically. Right now it's pockets of effort in a massive problem.