Procurement, SOWs, and Lawyers: Ideals Are Important
March 30, 2026
Another dinobaby post. No AI unless it is an image. This dinobaby is not Grandma Moses, just Grandpa Arnold.
I get a kick out of the posts on Marcus on AI. I try to stay somewhere near a cave in the great AI fights. Actually I hide out in my basement in rural Kentucky with some copper mess decorating portions of my little habitat for dinobabies. I covered the two windows so curious drones cannot easily spot me staring into space or just looking like any other drooling 82 year old.
Hey, Venice.ai, you did not tell me that this prompt violated your decency guardrails. What a surprise! Definitely just okay, but that’s the norm now, isn’t it?
Once in a while I read essays like “Is the US Military Actually Afraid of Claude? A New Theory of Why Anthropic Was Labeled a Supply Chain Risk.” I think the “reason” is interesting. I am not convinced, so let’s look at the possible cause.
The essay states:
Everyone in the industry realizes that LLMs without guardrails are an uncontrollable menace.
Would an uncontrolled AI cause a problem for warfighters? My initial reaction is, “Not if the AI can knock the objective off a checklist.”
The essay states:
If you think it is bad to use hallucinating LLMs to autonomously choose targets (I certainly do), you shouldn’t use any of them. Again this is in no way whatsoever special to Anthropic. Either all the LLMs are supply chain risks, or none of them are.
I urge you to read the full essay. I want to ask you, “Do you spot the flaw?” (That’s a snappy line I appropriated from a real estate sales dude with a trendy haircut. Thanks, Arvin!) In the first quote contains the word “everyone.” Obviously that’s not true because the US government is not in that set. Okay, logical fallacy number one. In the second quote, look at the word “either” or “none.” Yep, two more categorical affirmatives. Obviously there are some professionals in the US government not in these two sets.
Do these logical errors undermine the write up’s main point? I think most people won’t notice or just don’t care. Here’s my candidate for the core of the essay’s argument:
But the selective supply chain risk arguments just don’t hold water.
I think it is clear that some US government professionals do perceive Anthropic and the currently hot Claude as a problem. “Anthropic” in this context means the humans identified as out of line in the orderly world of the military world. And “Claude” is just the software system available to those in the US government who follow government rules quite closely. (No fair using Claude on your personal account from home to get that table fixed up for the presentation at the NRC tomorrow, please.)
The essay’s logic does not match up to the way governments work worldwide.
Here’s my view of the issue.
First, governments operate on a simple principle: We pay. You obey. When this precept is ignored for whatever reason, procurement has failed. Fixing up a grave procurement issue is a great deal of work. Some of that work slops across fiscal years and dribbles into different agencies, quasi government operations, and the often vilified consultants performing what some might call “real” work for the government.
Second, when a vendor not do what’s in the statement of work, more big problems manifest themselves. Re-competes can lead to objections from other vendors. If the other vendors are not happy with the outcome of the “hearing,” then the legal eagles take flight. The core idea is to contract with an outfit willing and generally believed to be able to “do the work.” Once it becomes clear that there is friction in the process, costs, time, and careers can be tossed on a bonfire of bureaucratic documents. This is not good. Most of those with some government mileage on their Hummer want to avoid big fires.
Third, the technical functionalities of any digital service are understood by a very small fraction of the “deciders.” I associate this word with a previous administration to whose tune I danced for a number of years. But I heard the same tune in other countries. This means that catchphrases and colorful, often misleading metaphors, are used as code to facilitate the decisions the deciders converge upon. Technology is rarely aligned with generalizations and snappy wordsmithing.
Why is Anthropic a poster child for smart software in the government? Answer: The company did not do the “we pay. you obey” thing. Second, the company found itself ensnared in the DEI – lefty catchphrase. Third, someone figured out that Claude could be and probably was what some in the US government perceived as “rogue code.”
Keep in mind I am aligned with the argument in the essay. I believe, however, that my comments provide some context for why a fairly straightforward procurement is going to gobble time and money to resolve. Keep in mind that governments worldwide operate the same way. I see no big change in the business processes of government. Therefore, the zippy technology outfits should just get with the program. Bid, win, and do. Notice I did not mention “push back.” Just a thought.
Stephen E Arnold, March 30, 2026
Comments
Got something to say?
-
- Subscribe to Beyond Search
Feature archive
News archive
-
