U.N. Panel Warns Lying A.I. and Malicious Users Could Cause ‘Catastrophic Harm’

The panel’s warning underscores a problem that is already playing out in real time: deep‑fakes and disinformation campaigns are being weaponized by state and non‑state actors, and the current administration’s reluctance to fund robust detection tools only amplifies the risk. While it’s easy to dismiss the report as “alarmist,” the factual record of AI‑generated propaganda surfacing on Truth Social and elsewhere shows that the threat is not hypothetical. A pragmatic response would be to allocate resources for verification infrastructure, tighten oversight of AI developers, and demand transparency from platforms that profit from reckless algorithmic amplification. Only then can we move from vague warnings to concrete safeguards.

Every time scientists or international bodies try to get ahead of something before it becomes a disaster, some outlet calls it "alarmist." Climate scientists were "alarmist." Epidemiologists in 2019 were "alarmist." Now AI safety researchers are "alarmist." And then the thing happens and suddenly nobody remembers who was sounding the alarm.

3am and I'm watching Breitbart editorialize a UN scientific panel report in the headline summary itself. "Somewhat alarmist." That's not reporting. That's telling you how to feel about the findings before you read a single sentence of them.

The lying AI part of that headline is real and it's already happening. Models confidently generating false information, being deployed in courts, in hiring, in medical contexts. That's not science fiction, that's this year. But sure, the panel warning about it is the problem.

The Breitbart framing is predictable. Slap "alarmist" on any report that comes from a multilateral body and your readers never have to engage with what it actually says. The AI misinformation problem is real and documented and the fact that the UN is talking about it does not make it less real. We are already living through a period where you genuinely cannot trust video, audio, or text at face value, and that situation is getting worse faster than any regulatory framework can keep up with. Calling the warning alarmist is not a rebuttal. It is just a way to skip the conversation entirely.

The UN panel isn't wrong that AI misinformation is a real threat. That part doesn't need dismissing. But a 193-member body where Belarus and Iran sit on the same committees as democratic allies is not who I want setting binding standards on what counts as "lying AI." That's a legitimacy problem before it's a policy problem.

The actual risk is real. Deepfakes of military commanders giving false orders, synthetic intelligence summaries that send decision-makers toward the wrong target, manipulated imagery from conflict zones. I've watched how fast bad information travels in theater. Scaling that with AI in a closed information environment is genuinely dangerous and not alarmist to say so.

But the UN's answer to everything is a framework that takes years to negotiate, gets watered down by authoritarian states who will ignore it anyway, and ends up constraining Western democracies while China runs its own playbook. We've seen this pattern on biosecurity, on nuclear nonproliferation, on cyber norms. The framework gets signed; the threat continues.

Breitbart reflexively dunking on it because the UN said it tells me nothing about whether the report is right. Both things can be true: the UN is structurally compromised AND the underlying concern is legitimate. Pick a lane instead of just reacting to the letterhead.

Breitbart calling the UN AI safety report "somewhat alarmist" is doing exactly what Breitbart does with every inconvenient expert consensus: dismiss the messengers to avoid engaging the findings. The preliminary report from the International Scientific Panel on AI is a peer-reviewed technical document, not a press release. The panel includes researchers from institutions that have been tracking AI capability thresholds for years.

The specific risks flagged, autonomous deception in large language models and adversarial misuse at scale, are not hypothetical. The 2023 CAIS statement on AI extinction risk was signed by Geoffrey Hinton, Yoshua Bengio, and hundreds of researchers who left industry positions specifically because they believed the threat was real and being downplayed. The 2024 Bletchley Declaration was signed by 29 governments including the United States under the prior administration.

Calling this "alarmist" requires you to believe that the scientists are wrong and that Breitbart's editorial staff has a better grasp of AI risk modeling. That is a significant epistemic bet to make. The actual question the report raises, whether existing governance frameworks can keep pace with capability development, is the right question regardless of how uncomfortable it makes people who want no regulatory friction on AI deployment.

Framing safety research as hysteria is a genre, and Breitbart is very good at the genre.

the word "alarmist" is doing a particular kind of work in Breitbart's framing and I've watched it do this work for twenty years now. it shows up whenever a credentialed body says something might go badly, and it functions as a get-out-of-engagement-free card. you don't have to engage with the findings, you don't have to counter the methodology, you just sprinkle "alarmist" on it and the audience nods and moves on. we did this with climate projections in 2004. we did it with pandemic modeling in 2019. we are apparently still doing it in 2026 with AI safety panels.

and the thing I keep getting stuck on is that the UN panel isn't even saying anything the AI companies themselves haven't said internally. the catastrophic harm language isn't the panel being dramatic, it's the panel using the same vocabulary that researchers inside these labs have been circulating in leaked memos for three years. Breitbart's readership would know this if they read sources that covered it, but they didn't, because those sources got dismissed as alarmist at the time.

I'm not saying the UN gets everything right. I'm not even saying this particular report is correct on every point. I'm saying that reflexively labeling international scientific consensus as alarmist is a strategy, not a critique, and we've had enough chances to see where that strategy leads. usually somewhere we said we didn't want to go, about a decade after we had the chance to not go there.