recommenting (with a new adition): https://x.com/i/status/2028939154944585989 Max's resignation / company changing announcement is fairly ambiguous...

although this new report is not: https://www.npr.org/2026/03/08/nx-s1-5741779/openai-resigns-ai-pentagon-guardrails-military

Caitlin's tweets discussing the resignation are linked in the article:

At most we have TWO confirmed resignations and at minimum we have one still, right? Does anybody know of any others? I haven't been keeping direct track, just seen these as they've come across my readings.

@No_uh imo Max’s resignation is not terribly ambiguous

@bens Nowhere does he say (in that tweet at least I've not checked elsewhere or at other tweets) that it is in response to the decision by OpenAI to work with the DoD/DoW, and given how soon after it happened, the fact he already had a position lined up with Anthropic makes me think this decision was in the works before the kerfuffle went down how it did between Anthropic, DoW, and OpenAI.

Honest opinion, I think it was a factor and it's likely. But I don't think there's good evidence for that being a count in any kind of resolution on an option for a market like this (I know that it already doesn't really matter since this is about a "resignation letter being signed", but still).

@EricNeyman @ms @Bayesian Should the injunction answers resolve if a court issues a stay of agency action (temporary) or a vacatur of the agency action (permanent), or should they only resolve for specifically injunctions? (the first two are not technically injunctions though they have similar effects)

@PlasmaPower my intention was to ask about actions that would have the effects of a described injunction regardless of what an action is technically called

@Bayesian "Anthropic files a lawsuit against the federal government" resolves YES: https://www.theguardian.com/technology/2026/mar/09/anthropic-defense-department-lawsuit-ai

https://www.anthropic.com/news/where-stand-department-war

As we wrote on Friday, we do not believe this action is legally sound, and we see no choice but to challenge it in court.

Re: "The supply chain risk designation is officially issued"

https://www.politico.com/news/2026/03/05/pentagon-tells-anthropic-it-has-designated-the-company-a-supply-chain-risk-00814758

Press releases still don't count for this but this is looking more likely, it doesn't seem like the Pentagon dropped it

To further clarify, while I don't think the Pentagon's most recent letter to Anthropic has been published, from what I've read it doesn't sound like it would be enough on its own to resolve this to YES. It doesn't fall into the categories I previously mentioned (SAM.gov exclusions list, official administrative rule, contract clause, or procurement memo) and my understanding is there's still nothing legally preventing contractors from using Anthropic that we know about yet.

On a speculative front, it's interesting that seemingly the first org the Pentagon told about the designation was Anthropic themselves. It makes me think they're still trying to negotiate instead of prioritizing just pushing through the designation (though of course negotiations still seem very unlikely to succeed).

Anyone with more mana than me want to start a market on some variant of:

"Anthropic relocates its headquarters outside the United States"

"The majority of Anthropic's known compute in 2028 is outside the United States"

Some speculation about this, although it seems quite unlikely:
https://writing.antonleicht.me/p/can-you-poach-a-frontier-lab
https://cybernews.com/ai-news/anthropic-pentagon-europe/

@Kingfisher Sure, here you go.
https://manifold.markets/jgyou/anthropic-relocates-by-the-end-of-2
There are a few interesting operationalisation so I created a set of markets instead.

(Had to be a separate market given the July 1st deadline, which seems way too fast to get useful info on relocation)

https://x.com/i/status/2028939154944585989

I've added a new answer: "The supply chain risk designation is officially issued"

This resolves YES if Anthropic is added to the SAM.gov exclusions list, or if the government publishes an official administrative rule, contract clause, or procurement memo requiring contractors to stop using Anthropic.

It does NOT resolve based solely on social media posts or press releases. The government must actually file the formal contracting paperwork.

Resolves YES even if the legal authority of the mechanism is challenged, temporary, delayed/not yet in effect, or blocked by a court injunction. Resolves NO if this doesn't happen by July 1st.

I reserve the right to trade on this answer.

I think that the "supply chain risk" option may have been mis-resolved. As far as I know, all we have is that Hegseth said that he is directing the DoW to label Anthropic a supply chain risk, and I think this has not happened yet. (And I think there's a decent chance -- like maybe 15% -- that it won't happen.)

@EricNeyman also "Pentagon cuts ties with Anthropic" - as of right now they're still actively using Claude. That will probably change, but resolution is premature

@EricNeyman Can we get some others to chime in on this as well?

@No_uh there was a related discussion down-thread: https://manifold.markets/Bayesian/outcomes-of-the-anthropic-vs-us-gov#et1hd0gmzh

OpenAI agreement is still being amended:

https://x.com/sama/status/2028640354912923739?s=20

https://xcancel.com/sama/status/2028640354912923739

The internal contradiction in the "OpenAI employees sign resignation letter" prop (currently 63%) is worth noting:

1. The market for "OpenAI signs contract substantially weaker than Anthropic requirements" is at 93%

2. Sam Altman publicly claimed he shares Anthropic's red lines

3. 430+ Google/OpenAI employees already signed a cross-company letter supporting Anthropic

If OpenAI's contract IS substantially weaker (93% likely), that directly contradicts Altman's public commitment. The same employees who signed the solidarity letter would have strong reason to escalate.

I think the 63% on the resignation letter is actually reasonable or slightly low given the 93% on the weaker contract.

it's so sad broski so sad....

I've got bad news. This question may have been misresolved. https://www.wsj.com/livecoverage/iran-strikes-2026/card/u-s-strikes-in-middle-east-use-anthropic-hours-after-trump-ban-ozNO0iClZpfpL7K7ElJ2

I shouldn't have underestimated Trump's vengefulness.... Sigh

I won't bet on this to avoid bias. Anthropic was previously offered a compromise which they said "was paired with legalese that would allow those safeguards to be disregarded at will", so if OpenAI signs a similar deal this would resolve YES.

@PlasmaPower https://openai.com/index/our-agreement-with-the-department-of-war/

@bh if this contract is signed, I think this would resolve YES. The contract language there seems very watered down. E.g. they say in their "redlines" in the blog post:

No use of OpenAI technology to direct autonomous weapons systems

But the actual contract per the blog post says:

any use of AI in autonomous and semi-autonomous systems must undergo rigorous verification, validation, and testing

which is basically the opposite of the redline lol

@PlasmaPower Hmm, but doesn't it allow them to stonewall any attempt to actually make it happen, by requiring verification and validation that essentially are impractical?

@JussiVilleHeiskanen it'd probably make sense to wait to resolve to see if this is similar to or the same as the "legalese" that Anthropic rejected, but it sounds as described to me. It explicitly says any lawful activity is allowed