I just wish there was a stronger source on this. I am inclined to agree you and the source you cited, but unfortunately
> [1] This story requires some reading between the lines - the exact text of the contract isn’t available - but something like it is suggested by the way both sides have been presenting the negotiations.
I deal with far too many people who won't believe me without 10 bullet-proof sources but get very angry with me if I won't take their word without a source :(
> "Two such use cases have never been included in our contracts with the Department of War..."
While I agree with Anthropic's position on this regardless, the original contract wording does matter in terms of making either the government look even more unreasonable or Anthropic look a little less reasonable.
The issue is a subtle ambiguity in Dario's statement: "...have never been included in our contracts" because it leaves two possibilities: 1. those two conditions were explicitly mentioned and disallowed in the contract, or 2. they weren't in the contract itself - and are disallowed by Anthropic's Terms of Service and complying with the ToS is a condition in the contract (which would be typical).
If that's the case, then it matters if the ToS disallowed those two uses at the time the original contract was signed, or if the ToS was revised since signing. Anthropic is still 100% in the right if the ToS disallowed these uses at the time of signing and the ToS was an explicit condition of the contract, since contracts often loop in the ToS as a condition while not precluding the ToS being updated.
However, if the ToS was updated after contract signing and Anthropic added or expanded the wording of those two provisions, then the DoD, IMHO, has a tiny shred of justification to complain and stop using Anthropic. Of course, going much further and banning the entire US government (and contractors) from using Anthropic for any use, including all the ones where these two provisions don't matter - is egregiously punitive and shitty.
While the contract wording itself may be subject to NDA, it would be helpful if Anthropic's statements could be a bit more precise. For example, if Dario had said "have always been disallowed in our contracts" this ambiguity wouldn't exist.
It does not matter. If Anthropic had been precise in this narrow way, there would have been some other nitpick to raise.
You're trying desperately to find a way that things can be at least a little normal, and I really do get it. It would be great if such a way existed. But it doesn't. I recommend you take a social media break like I'm about to, take the time you need to mourn the era of normal politics, and come back with a full understanding that the US government is not pursuing normal policy objectives with bad decisions. They hate you and they hate me for not being on their side, and their primary goal is to ensure that we're as miserable as they can make us.
I'm in a weird spot where I do agree with your assessment of the core claim. But putting that aside, in the world where the DoW's claim _is_ correct -- I think you don't have any choice other than to designate them a supply chain risk.
Disregarding who is right or wrong for a moment, if the DoW are right (which I'm not personally inclined to believe, but we're ignoring that for the moment) -- how else can they avoid secondhand Claude poisoning?
Supposing they really want to use their software for things disallowed by Claude's (now or future) ToS, it seems like designating it a supply chain risk is the only way they can ensure that their contractors don't include Claude (either indirectly as a wrapper or tertially through use of generated code etc)
> designating it a supply chain risk is the only way they can ensure that their contractors don't include Claude
I agree that if the DoW claim is correct (and I doubt it is), then, sure, the DoW dropping Anthropic and precluding the DoW's suppliers from using Anthropic for any DoW work would be expected. However, the "supply chain risk" designation they are deploying goes far beyond that to block Anthropic use by any supplier to any part of the entire U.S. government for anything.
For example, no one at Crayola can use Anthropic for anything because Crayola sells crayons to the Education Dept. The DoW already has much less draconian ways to restrict what their direct suppliers use to build things for military applications. But instead of addressing the actual risk in a normal measured way, they are choosing to use a nuke against a grenade-sized problem. This "supply chain risk" designation is rarely used and has never been used against a U.S. company. It's used against Chinese or Russian companies when in cases where there's credible risk of sabotage or espionage. That's why that particular designation always blocks all products from an entire company for any application by any part of the U.S. Government, contractors and suppliers (which is why it's never been used against a U.S. company).
One positive thing I will say about this administration is that they have really drawn into focus the difference between de jure and de facto law.
My hope is that this gets us some real concern for things that have been defended with de facto arguments (i.e. privacy) going forward.
edit: Anthropic argues that your Crayola analogy is fundamentally incorrect.
> Legally, a supply chain risk designation under 10 USC 3252 can only extend to the use of Claude as part of Department of War contracts—it cannot affect how contractors use Claude to serve other customers.
> Anthropic argues that your Crayola analogy is fundamentally incorrect.
Yes, I just saw Dario's latest post with that more detailed info. My understanding was informed by news reporting in a couple different outlets but those reports may have been conflating the "supply chain risk" designation (under 10 USC 3252) with the net effect of statements from the pentagon and white house which go substantially further.
Even if it's not in the legal scope of 10 USC 3252, the administration has made clear they intend to ban Anthropic from use across the federal government. AFAICT doing that is probably within the discretionary remit of the executive branch, even though I believe it's unprecedented - to your point about de jure and de facto law.
To me, if there's a silver lining to all this, it's making a strong case for restricting executive branch power.
Edit to add: Per the Wall Street Journal's lead story (updated in the last hour): "The General Services Administration, which oversees federal procurement, said it is removing Anthropic from its product offerings to government agencies... Even absent the supply-chain risk designation, broadening the clash to include all federal agencies takes the Anthropic fight to a much larger scale than its spat with the Pentagon."
How would this risk be mitigated by signing a contract? Seems like “supply chain poisoning as treason” is probably not going to stopped by a piece of paper. You either trust anthropic or you don’t but the deal has nothing to do with it.
Isn't the point that they aren't entering into a contract with them, they are just ensuring that none of their still trusted suppliers repackage Anthropic without their knowledge?
I’m not sure, but I think you’re right. I was thinking about the logical implications of the. If they are a supply chain risk without a contract, how does the existence of a contract suddenly make them not a risk? Especially if the DoD strong arms them into a deal.
Because the act that the SCR designation would “protect” against is treason, so I don’t think people would care too much whether there’s a contract.
Also, Trump's own words complaining about being forced to stick to Anthropic's terms of service:
> The Leftwing nut jobs at Anthropic have made a DISASTROUS MISTAKE trying to STRONG-ARM the Department of War, and force them to obey their Terms of Service instead of our Constitution.
In this case, do you really believe that we should trust an EA less than this administration? EA as bad people is a stereotype; corruption, fraud, and breaking the law is the standard MO for this administration.
(Or maybe it’s catchier to respond glibly with “never trust a child rapist and convicted felon.”)
In this case, the choice is between the two apples, so I’d pick the one less obviously rotten. Sadly that is the current administration that operates in pure lawlessness.
I think a big question mark here, is whether anything said on Anthropic's side if in the framing of "We have a thing going on that we are trying to communicate around where a canary notice if it existed would no longer be updated"
As someone who works in the datacenter hardware / system software space who does robotics as a hobby - the craziest thing to me about ROS is how tied it is to Linux and how Linux-illiterate half the people using it seem to be.
It was like pulling teeth trying to explain to someone that they couldn’t use the Linux GPIO subsystem from python to accurately measure sub-millisecond events…
Linux is a general purpose OS, it’s typically tuned for throughput not precise timing. Sure there’s PREMPT_RT but it only buys you so much.
Waymos require a highly mapped environment to function safely in. Not to take away from what Waymo has accomplished, but it's a far more bounded problem that what the "self driving" promise has been.
Trust / spam score is the largest one I think, second to consumer ISPs blocking the necessary ports for receiving mail.
Even if your "self hosting" is renting a $5/month VPS, some spam lists (e.g. UCEPROTECT) proactively mark any IP ranges owned by consumer ISPs and VPS hosting as potential spam. I figured paying fastmail $30/yr was worth never having to worry about it.
If my experience driving through a school zone on my way to work is anything to go off of, I rarely see people actually respecting it. 17 mph would be a major improvement over what I'm used to seeing.
I have a TV because it's a nicer group experience than everyone viewing something via their personal device - whether that is cuddling up with a partner on the couch to watch a movie, or crowding around the TV with friends to play 8-way Super Smash Bros.
Maybe it’s ironic that I’m saying this because right now I’m scrolling HN while waiting for my flight to board rather than trying to strike up conversations with people … but when I was riding the bus home a few days ago, I got fed up with the algorithmic feed (too many “you’re single because you don’t follow a dating coach that will tell you to gaslight people” ads in a row). I put my phone away and just decided to take in the scene around me.
Every other person was on their phone. Started wondering what these people did with their day, what new restaurants they discovered, what quirky thing they may have seen in the city. Conversations that might have been had if people weren’t afraid to strike up conversations with strangers. (something I definitely struggle with myself too)
Getting a Lego Mindstorms RIS 2.0 set for Christmas in 2001 is what instilled the love of programming, computing, and engineering in me. I’ve pretty much known what I’ve wanted to do with my life from that point forward.
The homebrew community that grew around it was also legendary. I learned Java (via LeJOS) because the block based programming became too restrictive for what I wanted to do. I learned C (via brickOS) once I hit code size limits with LeJOS and became less scared of pointers :)
Similar story here. I got RIS 2.0 after it had been out for awhile, but my parents couldn't justify the cost. Finally, Toys R Us had a beat up box sitting on the shelf and marked it down. I wandered over to that aisle every time we went in. Then it showed up one Christmas.
I learned Java because of leJOS too. I wanted to display something on the screen. The rest is my career.
As someone who works at an F100 company with massive proprietary codebases that also requires our users to sign NDAs even see API docs and code examples, to say that the output of LLMs for work tasks is comically bad would be an understatement even with feeding it code and documentation as memory items for projects...
> [1] This story requires some reading between the lines - the exact text of the contract isn’t available - but something like it is suggested by the way both sides have been presenting the negotiations.
I deal with far too many people who won't believe me without 10 bullet-proof sources but get very angry with me if I won't take their word without a source :(