"A 20-year-old man was arrested on Friday after throwing a Molotov cocktail at the San Francisco home of Sam Altman, chief executive of the artificial intelligence company OpenAI, according to the company and the police.

The incendiary device lit a fire on the exterior gate of Mr. Altman’s home before dawn, the San Francisco Police Department said in a statement. The suspect then fled on foot but was found about an hour later, at the OpenAI headquarters about three miles away, where he was threatening to burn down the building, the police said."

-NYT

I don't think the aggressors attacking Iran's facilities was due to "opposing AI development" in the way Eliezer talks about in his TIME article. This was not done with any concern for x-risk from AI, it was simply to hit Iran's capabilities.

I personally would consider it a misresolution if the strikes in Iran are used as the basis for YES.

Edit: either that, or the title doesn't align with the description.

@BrunoParga As far as I can tell from the question description, perceived x-risk is not a deciding factor here.

Question says:
> If an AI-aligned datacenter is attacked, this question resolves YES. If not, NO.

Multiple news sources seem to satisfy all elements of this condition:
- It's a datacenter
- It's AI-aligned (from news below: "forming the backbone of the country’s AI network", "provides infrastructure to the country’s AI capabilities", "underpinning the country’s national artificial intelligence platform")
- It was attacked

Which part would you disagree with?

@Kingfisher the title and the mention to the Eliezer article make this not super clear-cut in my view.

@BrunoParga I'm thinking about this. It does seem clear that this wasn't an attack because of X-risk, I agree with you there, and it's not the sort of thing EY was taking about.

But that isn't what I asked. It certainly was attacked to oppose AI development, just specifically Iranian AI development. I moved far away from what EY said otherwise. For instance, to resolve YES, I wouldn't need an airstrike, or for the data center to be "rogue"- a mob of people breaking into an Openai data center with axes because they hate GPT-5 would definitely resolve yes. Even if they didn't care about Google Gemini, or whatever. It's hard to resolve based on discerning underlying intentions...

Anyway I would appreciate reading further arguments if anyone has them

@TiredCliche I hear you. I think the title and Eliezer quote make the x-risk interpretation reasonable, such that a YES now means traders would be better off if they were unfamiliar with that context.

I think it is suboptimal to punish traders for information they possess; prediction markets should encourage, not discourage, information (and also meticulous question reading and writing). Still, now I wouldn't go so far as calling YES a misresolution 🙂

@TiredCliche Yesterday's U.S.-Israeli airstrike on Sharif University's datacenter causes this to resolve YES, I believe:

- CNN: https://www.cnn.com/2026/04/06/world/live-news/iran-war-us-trump-oil?post-id=cmnnhgyrk0000356s3jf35no3 "an airstrike targeted Tehran’s Sharif University of Technology on Monday morning, the semi-official Tasnim News Agency reported.

The strike damaged the university’s computing center and GPU facility, which provides infrastructure to the country’s AI capabilities, according to Tasnim, which is affiliated with Iran’s Islamic Revolutionary Guard Corps (IRGC)."
- https://www.aa.com.tr/en/middle-east/us-israeli-jets-bomb-artificial-intelligence-infrastructure-in-iranian-capital/3894678 "A US-Israeli strike targeted a data center at Sharif University of Technology in Tehran"
- https://gvwire.com/2026/04/06/us-iran-study-ceasefire-plan-as-deadline-nears-on-trumps-hell-threat/ "A U.S.-Israeli attack hit the data center at Sharif University of Technology in Tehran, damaging infrastructure underpinning the country’s national artificial intelligence platform and thousands of other services, Fars News Agency said on Sunday."
- https://breakingthenews.net/Article/Strike-on-Iran%27s-tech-university-damaged-data-center/66014093
- https://www.news18.com/world/iran-threatens-strike-us-ai-centre-abu-dhabi-openai-stargate-data-centre-iran-missiles-warning-sharif-university-attack-ws-l-10018399.html "[the semi-official Tasnim News Agency] mentioned the attack in Tehran damaged the computing centre and GPU facility at Sharif University of Technology"

'Within Missile Range': Iran Threatens To Target US-Linked AI Centre In Abu Dhabi Amid Strikes

Iran has warned that a US-linked artificial intelligence facility in Abu Dhabi could be considered a strategic target, following an attack at Tehran’s Sharif University.

@Kingfisher how big was the “GPU facility” and how critical to the country’s AI though? Would it have been targeted anyway?

@mariopasquato Exact size (GPU count/type, total power) are not public, from what I can tell - not that this question specified a hard threshold we could compare against. However:

- Reporting says "The facility underpins the country’s artificial intelligence platform along with thousands of other digital services" https://www.aa.com.tr/en/middle-east/us-israeli-jets-bomb-artificial-intelligence-infrastructure-in-iranian-capital/3894678
- https://thecradle.co/articles/us-israeli-strikes-target-irans-most-prominent-tech-university-target-ai-infrastructure#:~:text=The%20facility%20supports%20thousands%20of,of%20the%20country's%20AI%20network. "The facility supports thousands of integrated systems and research projects, forming the backbone of the country’s AI network."

It was widely considered the most advanced AI facility in the country.

How about cyberattacks? In the spirit of the quote, I'm talking more about physical attacks. So if it's a ransomware, even if Sam Altman has to pay $10 million [KROMER] to get his password changed back, that wouldn't count.

But it the goal of the cyberattacks would be making all the servers overheat and catch on fire, then I would resolve YES! That would have to at least sorta happen though. I wouldn't count someone online saying they tried to do Stuxnet 2

How would this resolve in the context Eliezer actually wrote it - to enforce an international treaty on AI?

Nuclear facilities in, say, Iran are targeted more or less regularly, so it's not like there's a strong international norm against this.

@BrunoParga It would resolve YES.

Probably need to specify what counts as an attack. A certain $ value of damage maybe? Just wondering if some minor vandalism as part of a protest would count or if we're just talking air strike level damage.

@Alfie yeah I think this needs to be defined!

@Tomoffer Yeah, I'm talking about a serious attack, with a serious dollar amount level of damage.

However LLM companies are kinda notorious for making up numbers, so I don't think I'll be trusting what they say. It doesn't have to be air strike level damage- a mob of people breaking into the server rooms and destroying them by spraying fire extinguishers would definitely suffice.

@Tomoffer But I also would consider it to resolve yes if the intentionality is there. If the mob makes it in before being mowed down by Gatling guns I would still resolve yes.

@TiredCliche What about a lone wolf attack, my least favorite type of wolf attack? In my view, it actually has to have a possibility of destruction for the act to be valid. A person shooting at a building would not resolve yes. It might destroy something. Someone might die. I don't want to dismiss those things.

But this market is about a serious attack against a data center that could meaningfully destroy its capabilities in some way.