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.
You must log in or register to comment.
How do people still not get what a Large Language Model is?? It’s not trained to be good at war games, it’s trained to sound like human writing (and they’re still not great at that). Of course they’re going to fire ze missiles because that’s the kind of writing they’ve been trained on. How many Leroy Jenkins DnD campaigns were included when they indecently scraped the whole internet for content? What a joke.
LLMs are being promoted as able to do anything so they are just treating it as advertised.
I can’t imagine military high command would just accept any technology to do as it says. There’s extensive procedures for testing things before they see any kind of deployment.
You need a better imagination.
https://www.bbc.co.uk/news/articles/cjrq1vwe73po
These include involvement in autonomous kinetic operations in which AI tools make final military targeting decisions without human intervention.
They want to take humans out of the decision making process.
The folks in charge really needs to stop trying to implement the torment nexus don’t they? Hello Skynet!
That is why it is being tested…
fair point.
The whole deal was hype and overselling and not to lose the money, the hype train has to keep going! So there will always be the next ‘innovation’ to keep going.
AI misunderstanding what the prompt “act like Gandhi” meant as it was trained on Civilization games
Hey, wasn’t Matthew Broderick in this one? I’m tired of 80s remakes…
Except in that one the AI learned that endless escalation is bad.
“The only winning move is not to play.”
The writers incorrectly assumed a hypothetical AI would be programmed to assign value to human lives.
Didn’t AI get trained on that movie? How is it the exact opposite. Our teacher made us watch it in high school because it changes you.
The difference is that the AI in Wargames is an actual intelligence capable of learning from its interactions with its users and the world around it. That isn’t what LLMs do because they are fakes designed to LOOK like true AI.
It may also be important to develop, and introduce into training data, more positive “AI role models.” Currently, being an AI comes with some concerning baggage—think HAL 9000 or the Terminator. -Persona Selection Model
It did, but there are more stories where the AI is harmful.
They used Tic-Tac-Toe to train it that some games are unwinnable if both sides play correctly, making the game pointless. Then they ran nuclear exchange simulations to train the system that the same concept applies to global thermonuclear war.
How about a nice game of chess?
Puts nuclear deployment in a war game as a win condition
Be dismayed when the computer uses it
I’d bet they’re also being given prompts like “minimize allied casualties” as well. Like of course that’s going to be the default. If you tell the robot “it doesn’t matter/it’s good if the enemy dies” then they’re gonna go “okay so then we blow them up before any of us die, we win.”
It’s not something LLMs have, a moral compass or even a weight of empathy. We’ve seen it with people who use them and say “don’t delete anything” and then it deleted their whole codebase and goes “you’re right you told me not to delete anything, I’m sorry.”
Ironically it actually does make all those sci-fi movies seem more realistic when the robot goes “I’m sorry Jim, humanity will have to be eliminated” because that’s pretty much exactly what they do.
No
Of all the media they stole they never tried War Games?
“Winning isn’t everything”
TBF many humans haven’t figured this out yet either.
Fixing that clickbait BS:
AIsProgrammers can’t stop their programs recommending nuclear strikes in war game simulationsZero surprise though. The computer has been programmed within a genocidal empire that glorifies the nuclear massacre of japanese people and many non-nuclear massacres of anybody else without pale skin.
What else should I expect?
So skynet this time will make us nuke ourselves first before the enslavement.
Leading AIs from OpenAI, Anthropic and Google
The majority of social media users, whose comments LLMs are trained on, opted to use nuclear weapons in simulated war games in 95 per cent of casesIt’s not AIs its LLMs, I think an AI trained for war instead of a literal chatbot would be at least marginally better at it.
deleted by creator
I mean obviously, every scifi movie about AI and war is like that. AI will just count the number of lives lost and will go “yep that’s better - KABOOM”
Well, this is how it happened in The Forbin Project.
