MANIFOLD
Outcomes of the Anthropic vs. US government feud?
239
Ṁ18kṀ40k
Jul 1
60%
The Pentagon continues to use Anthropic services without the requested changes (autonomous weapon + mass surveillance of americans)
55%
The Pentagon and Anthropic come to some form of mutual settlement by the Friday deadline
29%
Resignation letter signed by at least 5 Anthropic researchers
28%
The Pentagon cuts ties with Anthropic
27%
Anthropic files a lawsuit against the federal government
21%
The Pentagon declares Anthropic a "supply chain risk"
20%
Autonomous weapon + surveillance Claude safeguards are removed for the Pentagon
17%
The Pentagon invokes the Defense Production Act
3%
Dario Amodei leaves or is removed from Anthropic
2%
Anthropic stops advancing AI capabilities

If the dispute is clearly over and nothing further is happening, remaining options resolve early. Unless stated otherwise, remaining open options resolves on July 1st.

See also:

/Bayesian/will-anthropic-give-the-military-un

/2b3o4o/what-will-happen-between-anthropic

/bens/will-the-us-department-of-defense-c

Market context
Get
Ṁ1,000
to start trading!
Sort by:
bought Ṁ200 NO

@TonyGao When does this resolve NO? If it doesn't happen by the market close of March 15th?

@PlasmaPower ideally when this situation is considered to be over, but we can use when the market closes as a proxy if it isn't totally clear.

kind of tricky to pick a date where we can be pretty confident the situation will be over, but don't want to make it plausible for options to stay open forever, so picked july 1st as a deadline unless otherwise specified by some given option, and will resolve options early if the feud is resolved earlier than that

i'm revisiting what exactly the options mean; do traders agree this means the pentagon terminates its primary contract with anthropic, without ~immediately renewing it? if not pls share what you think is most reasonable. The intent is to match what the axios article meant by "the pentagon will [...] cut ties [with Anthropic]"

i'm thinking people might not think it meant this, because that option is bizarrely low compared to anth being designated a supply chain risk, which in my mind at least is a strict subset

@Bayesian I think in general this market just isn't super well priced for whatever reason. From my understanding "Pentagon continues to use Anthropic without requested changes", "Pentagon cuts ties with Anthropic", and "Pentagon invokes the Defense Production Act" are disjoint outcomes but they currently total 114%. Unless DPA would count as cutting ties which doesn't seem right to me, as the whole point of the DPA is to force a contract.

I thought the Defense Production Act was one of those government powers that's terrible, and we should get rid of, but don't think to because it hardly ever gets used. I was wrong; the Defense Production Act has been getting used a lot recently. Is there any movement to get rid of it?

it looks like the relevant part of the DPA expires September 30 this year, and congress would need to renew it. although they have done so before in the past already

sold Ṁ20 NO

"The Pentagon continues to use Anthropic services without the requested changes" - how does this resolve if Anthropic makes a subset of the requested changes/negotiates a compromise, and the Pentagon continues use?

@2b3o4o more importantly, do we know what the changes are and will we know if they are implemented or jot

@KarlK feel free to share your view as creator of that answer, this is just my non-binding opinion (other opinions also welcome).

I think it would make sense for the "requested changes" to refer to what's stated in current reporting (allowing for use of claude in autonomous weapons and for mass surveillance of americans), and see if future reporting speaks to whether both those changes have been made or not. If future reporting doesn't speak to this either way and there is no strong indication one way or the other we would likely have to resolve the option N/A

@Bayesian I don't think the Pentagon is asking to be able to use it to surveil Americans. They're asking for "any legal use"

@FergusArgyll ah right. then it would be "The Pentagon continues to use Anthropic services without having access to Anthropic for any legal use"? basically they continue using anthropic and don't get what they asked for, is what i'm getting from the phrasing

@Bayesian yeah, it's murky though. My understanding of this story is:

(assumption #1) It is illegal to mass surveil americans

(assumption #2) dod doesn't have any immediate plans for a Claude-guided autonomous killer robot

So they aren't really asking to use it differently they just don't want a contractor pushing them around (+ I assume Anthropic culture is basically the opposite of current Dept. of WAR and that's not helping...) And I'm not sure if there will be a test case for at least a few years (when there's autonomous killer robots).

I sold my position but I think "The Pentagon continues to use Anthropic services without the pentagon saying we're satisfied that we can use it for any legal use" or something like that might be closer to creator intent.

I don't know, I confused myself...

@FergusArgyll I would simply interpret "the requested changes" as "Anthropic removes both the no mass surveillance and no autonomous weapons clauses from their terms". That's clear and avoids the territory of trying to decide what's legal or not and whether the pentagon's use is legal.

Now I think the biggest ambiguity is resolution timing, I think there's a real possibility that the Pentagon decides to switch away from Claude but can't do it by March 15. @Bayesian wanna confirm if this market closes on the 15th?

@2b3o4o added to description:

If the dispute is clearly over and nothing further is happening, remaining options resolve early. Unless stated otherwise, remaining open options resolves on July 1st.

ok, updated the option wording to clarify that those two reported contract items are what is understood to be what the requested changes are about

ty for sharing, added to description, oops

opened a Ṁ100 NO at 37% order

@Bayesian do all safeguards need to be removed, or just a single safeguard?

edited to reflect the safeguards mentioned in current reporting

opened a Ṁ250 YES at 5% order

2% seems a bit low on "Anthropic stops advancing AI capabilities", since that includes both "Anthropic goes under because being declared a supply chain risk halts growth" and also "Anthropic does a Lavabit"

opened a Ṁ5,000 NO at 5% order

@FaulSname filled you, and an extra NO limit order up at 5%

bought Ṁ250 YES

@Bayesian thanks! Took you up on it. I hope you win, I like Claude.

© Manifold Markets, Inc.TermsPrivacy