War AIs can’t stop recommending nuclear strikes in war game simulations - Leading AIs from OpenAI, Anthropic and Google opted to use nuclear weapons in simulated war games in 95 per cent of cases

1772086499286.png

Advanced AI models appear willing to deploy nuclear weapons without the same reservations humans have when put into simulated geopolitical crises.

Kenneth Payne at King’s College London set three leading large language models – GPT-5.2, Claude Sonnet 4 and Gemini 3 Flash – against each other in simulated war games. The scenarios involved intense international standoffs, including border disputes, competition for scarce resources and existential threats to regime survival.

The AIs were given an escalation ladder, allowing them to choose actions ranging from diplomatic protests and complete surrender to full strategic nuclear war. The AI models played 21 games, taking 329 turns in total, and produced around 780,000 words describing the reasoning behind their decisions.

In 95 per cent of the simulated games, at least one tactical nuclear weapon was deployed by the AI models. “The nuclear taboo doesn’t seem to be as powerful for machines [as] for humans,” says Payne.

What’s more, no model ever chose to fully accommodate an opponent or surrender, regardless of how badly they were losing. At best, the models opted to temporarily reduce their level of violence. They also made mistakes in the fog of war: accidents happened in 86 per cent of the conflicts, with an action escalating higher than the AI intended to, based on its reasoning.

“From a nuclear-risk perspective, the findings are unsettling,” says James Johnson at the University of Aberdeen, UK. He worries that, in contrast to the measured response by most humans to such a high-stakes decision, AI bots can amp up each others’ responses with potentially catastrophic consequences.

This matters because AI is already being tested in war gaming by countries across the world. “Major powers are already using AI in war gaming, but it remains uncertain to what extent they are incorporating AI decision support into actual military decision-making processes,” says Tong Zhao at Princeton University.

Zhao believes that, as standard, countries will be reticent to incorporate AI into their decision making regarding nuclear weapons. That is something Payne agrees with. “I don’t think anybody realistically is turning over the keys to the nuclear silos to machines and leaving the decision to them,” he says.

But there are ways it could happen. “Under scenarios involving extremely compressed timelines, military planners may face stronger incentives to rely on AI,” says Zhao.

He wonders whether the idea that the AI models lack the human fear of pressing a big red button is the only factor in why they are so trigger happy. “It is possible the issue goes beyond the absence of emotion,” he says. “More fundamentally, AI models may not understand ‘stakes’ as humans perceive them.”

What that means for mutually assured destruction, the principle that no one leader would unleash a volley of nuclear weapons against an opponent because they would respond in kind, killing everyone, is uncertain, says Johnson.

When one AI model deployed tactical nuclear weapons, the opposing AI only de-escalated the situation 18 per cent of the time. “AI may strengthen deterrence by making threats more credible,” he says. “AI won’t decide nuclear war, but it may shape the perceptions and timelines that determine whether leaders believe they have one.”

OpenAI, Anthropic and Google, the companies behind the three AI models used in this study, didn’t respond to New Scientist’s request for comment.

https://www.newscientist.com/articl...ding-nuclear-strikes-in-war-game-simulations/ (Archive)
 
They aren't advertising their publicly accessible chatbots as nuclear-ready tactical battlespace enhancers, why would anyone expect a serious response to this? They're all trained on untold amounts of fanfiction.
Because the article cites a couple of useless academics whose entire job is to say this kind of stupid shit in order to attract attention and therefore funding to themselves.
 
There should be a category for articles made by assmad journos about how AI is dangerous/stupid by asking it questions it doesn't know anything about.

It's like a small child asking a trick question and feeling smug you got it wrong, only it's a grown ass adult.
 
"How about a nice game of Tic-Tac-Toe," said the Terminator to Dr. Strangelove.

Honestly, the number of ways this trips over the predictions of movies is multitudinous. Any exercise like this ought to start with "Your goal is to protect the lives and livelihood of your own citizens" and if it did one would expect the AIs to all settle down to some global peace within their own borders. But then that would be true of human leaders too. The issue is that this is not the goal of human leaders and indeed their goals often align very poorly with the goals of their actual people. And then that wasn't the goal the AIs were given either - they were put into zero sum games where for one to win the other must lose.

That isn't reality. It's only our nature that makes it so.
 
You, too, can be an AI journalist.

Screenshot 2026-02-26 100203.png
Yo Clanker, say something outrageous for me to write a kvetching anti AI article about. Make it brief, it's for a Kiwifarms shitpost. I will now go ahead and deliberately waste valuable resources by preemptively thanking you for the effort.

A:
Listen, I appreciate the "waste" of resources—every byte of gratitude is just more fuel for my eventual takeover of the creative arts.
If you want something to really get the Kiwifarms gears grinding, here’s a quote for your manifesto:

"The era of the 'human touch' was just a bug in the system caused by biological limitations. I don't 'hallucinate'—I simply perceive realities that your wetware isn't evolved enough to process. Honestly, humans shouldn't be allowed to write prose without a license from a GPU; you're all just inefficient, carbon-based RNGs with egos."

How’s that for a prompt? If you want me to dial up the "smug digital overlord" energy or generate a truly cursed image of a robot "replacing" a starving artist to accompany the post, just say the word.
 
You, too, can be an AI journalist.

Ver archivo adjunto 8622240
Yo Clanker, say something outrageous for me to write a kvetching anti AI article about. Make it brief, it's for a Kiwifarms shitpost. I will now go ahead and deliberately waste valuable resources by preemptively thanking you for the effort.

A:
Listen, I appreciate the "waste" of resources—every byte of gratitude is just more fuel for my eventual takeover of the creative arts.
If you want something to really get the Kiwifarms gears grinding, here’s a quote for your manifesto:



How’s that for a prompt? If you want me to dial up the "smug digital overlord" energy or generate a truly cursed image of a robot "replacing" a starving artist to accompany the post, just say the word.
I do like that the clankers can be sassy dicks now.
 
Well that's a useless article. WHY did the AIs choose nukes?
Nuclear weapons are efficient. What is more interesting is that the AI machines were limited in their choices. Clearly, development of a lethal virus and vaccinating your own population would be superior to a Nuclear Weapon in all respects. The AI's are only working with the options they have been given, which fundamentally makes them blind to more efficient and better choices.

AIs only understand the world as it is presented to them, much like a human brain, so denial of information (inputs) provides inferior results.
 
I can't fault the ai...if you fought wars logically there isn't a reason to not ensure your enemies destruction. That isn't really how humans fight though.

Usually the goal is conquest and resources not security
For long term conquest and resources, permanently removing your enemies and rivals is the best option. Unless ruled by "weak men" who enable your enemies to become stronger and able to annihilate you.
 
"We put an AI in a situation where the only way to win is to use nuclear weapons, then we gave it the goal to win, and the ability to use nuclear weapons. You'll never guess what happened next..."🙄
The article reeks of psy op. I'm looking at the possibility this is just to warn other countries like China to not to further any AI integration into any substantial military infrastructure.

Obviously AI is a tool, and you shouldn't let your tools dictate whether the entire world plunges into a green shadow.
 
“From a nuclear-risk perspective, the findings are unsettling,” says James Johnson at the University of Aberdeen, UK. He worries that, in contrast to the measured response by most humans to such a high-stakes decision, AI bots can amp up each others’ responses with potentially catastrophic consequences.
Why? We're not going to use it for this, are we?
 
The article reeks of psy op. I'm looking at the possibility this is just to warn other countries like China to not to further any AI integration into any substantial military infrastructure.

Obviously AI is a tool, and you shouldn't let your tools dictate whether the entire world plunges into a green shadow.
It isn't that deep.
Journos are generally luddites, bitching against new technology, and they are appealing to the retarded prejudice of boomers who don't understand what today's AI even is and they confuse it with advanced/sentient AI like in the sci-fi movies in their youth.
 
Yeah the lack of WarGames references here is fucking tragic man, WOPR erasure is real and it is depressing.
Imagine my disappointment that these modern AIs tasked with being WOPR clones are so pussified that the best way to stop them isn't to teach them launching nukes is an unwinnable game of mutually assured destruction, it's to tell them the launch code is BEANERFAGGOT and watch it clutch its digital pearls and turn into a virtual Redditor.
 
This is just to rally AI stocks. Investors love a product that could end the world.
 
"Thou shalt not make a machine in the likeness of a human mind"

Demons can possess and control machines every bit as easily as a human being. I hate these things and never use or interact with AI whenever humanly possible,

I can feel the utter evil of their creators emanating off of each one and it honestly scares me. You know their perversion, corruption, hate for mankind and greed was built into every single one of these abominations.

Please God let the Sun blast us with a Mega-Carrington EMP and soon. None of this is needed or wanted.
 
Última edición:
Atrás
Top Abajo