Headlines
On February 13, the Wall Avenue Journal reported one thing that hadn’t been public earlier than: the Pentagon used Anthropic’s Claude AI through the January raid that captured Venezuelan Chief Nicolás Maduro.
It mentioned Claude’s deployment got here by way of Anthropic’s partnership with Palantir Applied sciences, whose platforms are extensively utilized by the Protection Division.
Reuters tried to independently confirm the report – they could not. Anthropic declined to touch upon particular operations. The Division of Protection declined to remark. Palantir mentioned nothing.
However the WSJ report revealed another element.
Someday after the January raid, an Anthropic worker reached out to somebody at Palantir and requested a direct query: how was Claude really utilized in that operation?
The corporate that constructed the mannequin and signed the $200 million contract needed to ask another person what their very own software program did throughout a navy assault on a capital metropolis.
This one element tells you every thing about the place we really are with AI governance. It additionally tells you why “human within the loop” stopped being a security assure someplace between the contract signing and Caracas.
How large was the operation
Calling this a covert extraction misses what really occurred.
Delta Pressure raided a number of targets throughout Caracas. Greater than 150 plane have been concerned. Air protection programs have been suppressed earlier than the primary boots hit the bottom. Airstrikes hit navy targets and air defenses, and digital warfare property have been moved into the area, per Reuters.
Cuba later confirmed 32 of its troopers and intelligence personnel have been killed and declared two days of nationwide mourning. Venezuela’s authorities cited a loss of life toll of roughly 100.
Two sources advised Axios that Claude was used through the energetic operation itself, although Axios famous it couldn’t affirm the exact position Claude performed.
What Claude may even have performed
To grasp what might have been taking place, you could know one technical factor about how Claude works.
Anthropic’s API is stateless. Every name is impartial i.e. you ship textual content in, you get textual content again, and that interplay is over. There isn’t any persistent reminiscence or Claude operating repeatedly within the background.
It is much less like a mind and extra like an especially quick guide you may name each thirty seconds: you describe the state of affairs, they provide you their greatest evaluation, you grasp up, you name once more with new info.
That is the API. However that claims nothing in regards to the programs Palantir constructed on prime of it.
You may engineer an agent loop that feeds real-time intelligence into Claude repeatedly. You may construct workflows the place Claude’s outputs set off the subsequent motion with minimal latency between suggestion and execution.
Testing These Eventualities Myself
To grasp what this really appears like in follow, I examined a few of these eventualities.
each 30 seconds. indefinitely.
The API is stateless. A complicated navy system constructed on the API would not must be.
What which may appear to be when deployed:
Intercepted communications in Spanish fed to Claude for fast translation and sample evaluation throughout a whole lot of messages concurrently. Satellite tv for pc imagery processed to establish car actions, troop positions, or infrastructure adjustments with updates each jiffy as new photographs arrived.
Or real-time synthesis of intelligence from a number of sources – alerts intercepts, human intelligence reviews, digital warfare information – compressed into actionable briefings that will take analysts hours to supply manually.
educated on eventualities. deployed in Caracas.
None of that requires Claude to “resolve” something. It is all evaluation and synthesis.
However if you’re compressing a four-hour intelligence cycle into minutes, and that evaluation is feeding instantly into operational selections being made at that very same compressed timescale, the excellence between “evaluation” and “decision-making” begins to break down.
And since this can be a categorised community, no one outdoors that system is aware of what was really constructed.
So when somebody says “Claude cannot run an autonomous operation” – they’re in all probability proper in regards to the API degree. Whether or not they’re proper in regards to the deployment degree is a totally completely different query. And one no one can at present reply.
Hole between autonomous and significant
Anthropic’s arduous restrict is autonomous weapons – programs that resolve to kill with no human signing off. That is an actual line.
However there’s an infinite quantity of territory between “autonomous weapons” and “significant human oversight.” Take into consideration what it means in follow for a commander in an energetic operation. Claude is synthesizing intelligence throughout information volumes no analyst might maintain of their head. It is compressing what was once a four-hour briefing cycle into minutes.
this took 3 seconds.
It is surfacing patterns and proposals sooner than any human workforce might produce them.
Technically, a human approves every thing earlier than any motion is taken. The human is within the course of. However the course of is now transferring so quick that it turns into inconceivable to judge what’s in it in quick paced eventualities like a navy assault.When Claude generates an intelligence abstract, that abstract turns into the enter for the subsequent choice. And since Claude can produce these summaries a lot sooner than people can course of them, the tempo of all the operation quickens.
You may’t decelerate to consider carefully a couple of suggestion when the state of affairs it describes is already three minutes outdated. The data has moved on. The subsequent replace is already arriving. The loop retains getting sooner.
90 seconds to resolve. that is what the loop appears like from inside.
The requirement for human approval is there however the capability to meaningfully consider what you are approving will not be.
And it will get structurally worse the higher the AI will get as a result of higher AI means sooner synthesis, shorter choice home windows, much less time to assume earlier than appearing.
Pentagon and Claude’s arguments
The Pentagon needs entry to AI fashions for any use case that complies with U.S. legislation. Their place is actually: utilization coverage is our drawback, not yours.
However Anthropic needs to keep up particular prohibitions – no totally autonomous weapons and prohibiting mass home surveillance of Individuals.
After the WSJ broke the story, a senior administration official advised Axios their partnership/settlement was beneath overview and that is the explanation Pentagon said:
“Any firm that will jeopardize the operational success of our warfighters within the area is one we have to reevaluate.”
However paradoxically, Anthropic is at present the one industrial AI mannequin authorised for sure categorised DoD networks. Though, OpenAI, Google, and xAI are all actively in discussions to get onto these programs with fewer restrictions.
The actual struggle past arguments
In hindsight, Anthropic and the Pentagon may be lacking all the level and pondering coverage languages may remedy this subject.
Contracts can mandate human approval at each step. However, that doesn’t imply the human has sufficient time, context, or cognitive bandwidth to really consider what they’re approving. That hole between a human technically within the loop and a human really in a position to assume clearly about what’s in it’s the place the actual threat lives.
Rogue AI and autonomous weapons are in all probability the later set of arguments.
As we speak’s debate must be – would you name it “supervised” if you put a system that processes info orders of magnitude sooner than people right into a human command chain?
Last ideas
In Caracas, in January, with 150 plane and real-time feeds and selections being made at operational velocity and we do not know the reply to that.
And neither does Anthropic.
However quickly, with fewer restrictions in place and extra fashions on these categorised networks, we’re all going to search out out.
All claims on this piece are sourced to public reporting and documented specs. We now have no personal details about this operation. Sources: WSJ (Feb 13), Axios (Feb 13, Feb 15), Reuters (Jan 3, Feb 13). Casualty figures from Cuba’s official authorities assertion and Venezuela’s protection ministry. API structure from platform.claude.com/docs. Contract particulars from Anthropic’s August 2025 press launch. “Visibility into utilization” quote from Axios (Feb 13).
