Settings

Theme

Pentagon boasts of using AI to write reports mandated by Congress (1.5mil users)

arstechnica.com

78 points by FrustratedMonky 3 days ago · 60 comments

Reader

dbvn 3 days ago

The reports will be just as useless as they were before

  • PunchyHamster 3 days ago

    We might have some disrepancies leaked just because AI noticed it and nobody redacted it out of output coz nobody read it before sending

SoftTalker 3 days ago

And Congressional staffers will be using AI to summarize the reports, no doubt.

  • gmerc 3 days ago

    Meaning the techbros are basically carrying taxpayer money out of the door in buckets.

    Happened at US aid too. Cut vaccine and food budgets and gave them billions in AI money to shovel into Grok and ChatGPT.

    It’s all a massive heist

FrustratedMonkyOP 3 days ago

The shocking part of the story is the scale.

1.5 Million Users just within the Pentagon?

"The Pentagon has made AI tools, starting with Google Cloud’s Gemini for Government, widely available to members of all six military branches through the department’s bespoke GenAI.mil platform since December 2025."

"The number of Department of Defense personnel using commercial AI tools such as Gemini through GenAI.mil has significantly increased from just 80,000 in December 2025 to 1.5 million in June 2026, the Pentagon CTO claimed during his remarks at the Hudson Institute."

  • not_a_bot_4sho 3 days ago

    "The Department of Defense is the country’s largest employer, with more than 2.1 million Military Service members and over 811 thousand civilian employees."

    From https://comptroller.war.gov/Portals/45/Documents/afr/fy2024/...

    • tananaev 3 days ago

      Presumably the whole point of AI is that we can start reducing that number. I am sure there's a significant number of people just working on various administrative tasks processing data and documents.

      • thatguy0900 3 days ago

        That's exactly what I want from the Pentagon, less real humans involved and more automated systems that don't even have the concept of morality and the social responsibility to be a potential whistleblower

        • jimt1234 3 days ago

          Worked fine in The Terminator. Searching the the nasdaq for "Cyberdyne Systems" to add to my portfolio.

      • mpalmer 3 days ago

        Unless we're doing away with human accountability, the responsibility to accuracy (and whatever other statutory requirements) will remain the same, it will just be concentrated among significantly fewer federal employees.

      • iAMkenough 3 days ago

        Hell yeah let’s take the largest employer in the country and significantly eliminate jobs. Great for the economic jobs report and economy as a whole. I’m sure you’ll get a check in the mail with the savings.

        They still won’t pass an audit, ever.

        • mynameisbilly 3 days ago

          Imagine if we had a federal jobs program for building high speed rail all across the country. sigh

          • iAMkenough 3 days ago

            Instead we have a federal jobs program for infiltrating States without their consent and performing Kavanaugh stops (where the SCOTUS says the government can detain you for DAYS if you're not white and look "suspicious" to an inadequately trained ICE or other federal agent engaged in domestic terrorism).

            Trump supporters are going to really appreciate when the "you're not white" part is dropped.

  • gmerc 3 days ago

    How do you think OpenAI is going to become profitable? GovGPT.

  • BurningFrog 3 days ago

    Everyone is starting to use AI. I use three AIs daily. Why would the US military be different?

    • otikik 3 days ago

      Well, you are not the Pentagon.

      I presume they might view their internal data being used to train a private company's training sets as a concern.

    • FrustratedMonkyOP 3 days ago

      I think, assume. With the Military, with the goal of killing people, that there is assumption of more human judgment being involved. Beyond even the issue of a man-in-the-loop for targeting systems. But even generally, that these are big decisions that shouldn't be farmed out completely.

      • wildzzz 3 days ago

        For every job in the private sector, there's probably someone in the DoD doing the same thing. In 2005, the tooth-to-tail ratio (combat to non-combat roles) was 1:8.1.

        I'm guessing the vast majority of the AI usage is for things any typical office worker would use it for.

      • BurningFrog 3 days ago

        There are 3 million people working in the US military.

        Very few of them ever come close to killing anyone.

        Mostly they're white collar workers doing regular desk jobs, where AI is just as useful as in other industries.

        • vjvjvjvjghv 3 days ago

          A military is usually a big supply chain operation with an attached small war fighting branch

    • vitally3643 3 days ago

      How many of your tasks decide whether entire populations of human beings live or die?

Ancalagon 3 days ago

Guess the don’t need that budget increase after all with so much additional efficiency, right?

goldenshale 2 days ago

That's exactly what they should be doing. Reports to congress made by automated, neutral AI will probably be more honest, effective, and have a clear means of improving them over time with better prompting in comparison to whatever random people assigned to it produce. We should do this across the government, and there would likely be a dramatic improvement in efficiency and quality.

  • jfengel 2 days ago

    The reports are only as neutral and honest as the prompts given to them.

    If you want the reports to be honest, then the Congress would write the prompt. If the Pentagon writes the prompts, then it can keep re-asking the question until it finds a spin that it likes.

  • lenkite 2 days ago

    You mean like the KPMG's AI-written report praising AI that was withdrawn after hallucinated cases studies ?

    AI Slop Targeting already caused significant civilian casualties in the Hormuz War.

clickety_clack 3 days ago

A great way to undermine government would be to get them locked on Lines of Report (LoR) pushed to congress each month.

  • Georgelemental 3 days ago

    Oh, no, this would be an amazing policy. Almost certainly a better use of their time and resources than whatever horror they would concoct otherwise

FrustratedMonkyOP 3 days ago

Literally on the timeline for AI-2027.

https://ai-2027.com/

<edit> AI-2027, not Project 2027

  • cyanydeez 3 days ago

    ok, first, this isn't superhuman AI; this is slop production at scale. we could do these with markov chains decades ago.

    The only difference now is the slop looks critically better, but there's no quality accounting.

    • FrustratedMonkyOP 3 days ago

      In AI-2027, there are many stages before superhuman AI. Military dependence on AI was one of them. This dependence creates incentive to do anything to not slow down, including ignoring any safety concerns.

      • ameliaquining 3 days ago

        I read that prediction as being about using AI for mission-critical stuff, which this isn't really.

  • FrustratedMonkyOP 3 days ago

    Not sure why the downvote.

    Part of AI-2027, one of the early steps, was government dependance on AI for routine jobs. They become dependent on AI, and thus less willing to slow down or put on any guard rails. Because they can't live without it, they keep accelerating.

    • lelandfe 3 days ago

      Likely because they said Project 2027 which is something different

      • FrustratedMonkyOP 3 days ago

        Did not realize there was a follow on to Project 2025, called 2027.

        • lelandfe 3 days ago

          evokes something different/doesn't exist

          At minimum that link does not describe a project and does not use that Proper Noun

dnw 3 days ago

Similar to Paperwork Reduction Act, we may need Slop Reduction Act.

  • sidewndr46 3 days ago

    Which would mandate an AI summary be added to each document indicating the amount of AI used to generate it.

josefritzishere 3 days ago

AI slop reports from lazy, incompetent leaders? I'm shocked!

  • CGMthrowaway 3 days ago

    Half of those leaders are Democrats. They're not all lazy and incompetent.

    • steve_adams_86 3 days ago

      Is the implication that Democrats are less lazy and incompetent?

      Republicans haven't been lazy at all in their efforts to run the USA in my opinion. Democrats seem to have failed to match their energy. They've also failed to be competent enough to match their game.

      I'm not a fan of the Republican party and I never have been, but I'm not impressed by the Democrats either. From here in Canada it all looks like a chaotic mess, rife with corruption on both sides of the aisle.

      I don't understand the distinction here.

    • KetoManx64 3 days ago

      Both parties hate you and your liberties and have happily sold out your and your kids future to eternal debt

htx80nerd 3 days ago

if any (D) President did the same thing there would be 250 comments talking about how amazing this is. a true step into our future, etc. (R) man bad. everything (R) man does is bad. I know this cuz the Media and Experts tell me!

  • advisedwang 3 days ago

    You are getting mad at something you made up yourself

    • htx80nerd 2 days ago

      obama talked tough about the border repeatedly. including during a state of the union.

      obama deported tons of people.

      obama did the 'kids in cages' thing.

      do you recall any protest? mass media outrage about fascism and racism?

      hillary, bernie, etc, all talked tough about immigration back in the day. no one cared.

defmetrix 3 days ago

I have no problem with this. If the AI has access to the funding and schedule data, it will probably give a more honest answer that the humans. And in reality, nobody in congress is going to take the time to read the report anyways. They will just vote the way they are told.

arjie 3 days ago

This is wonderful. Many of these reports are makework paperwork. One even wonders if pushing for them is just taking a page from the CIA Sabotage Manual and applying it to us. Considering Congress members barely read the bills they’re voting on, it’s probably insignificant that this pointless paperwork is dispensed with.

When we finally end Environmental Impact Reports by generating them at scale with AI we will finally be able to escape this plateau of ossification.

I’m not particularly attached to bullshit being manufactured by human minds.

  • pstuart 3 days ago

    > Considering Congress members barely read the bills they’re voting on

    That seems like a good opportunity for AI to be used as a summary.

    And on other fun note, in many cases Congress does not even write the bill, their patrons do and have them pretend to represent it.

  • advisedwang 3 days ago

    If these reports truly are so bad, the law should be changed to stop requiring them. But, lawmakers aren't choosing to do that. Maybe there's actually some good reasons for them to exist.

Keyboard Shortcuts

j
Next item
k
Previous item
o / Enter
Open selected item
?
Show this help
Esc
Close modal / clear selection