[22715] in Privacy_Forum
[ PRIVACY Forum ] Script of my national radio report yesterday re
daemon@ATHENA.MIT.EDU (Lauren Weinstein)
Tue Dec 16 11:05:57 2025
Date: Tue, 16 Dec 2025 07:55:52 -0800
From: Lauren Weinstein <lauren@vortex.com>
To: privacy-dist@vortex.com
Message-ID: <20251216155552.GA4242@vortex.com>
Content-Disposition: inline
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Content-Type: text/plain; charset="us-ascii"; Format="flowed"
Errors-To: privacy-bounces+privacy-forum=mit.edu@vortex.com
This is the script of my national radio report yesterday on the new
Google/Pentagon Partnership for Gemini Military AI. As always there
may have been minor wording variations from this script as I presented
the report live on air.
- - -
So yes, this has quickly turned into a rather strange Pentagon story
but before we get further into it I want to answer a question I've
seen posed frequently: Is it the Department of Defense or Department
of War?
And this turns out to be a pretty easy one. Only Congress can change a
name like that for legal or statutory purposes, and Congress has not
done this. So the official legal name is still Department of Defense.
An executive order created a secondary name which is Department of
War, and that can be used in public communications and basically
various other non-statutory contexts. It isn't clear how much has been
spent changing various stuff to that secondary name, but apparently
estimates are that to do total rebranding to Department of War would
cost many billions of dollars and perhaps Congress doesn't want to
rush into that. I have seen some references to Department of War as
the "cosplay" name but I'm fine just saying the Pentagon so that's
what I'll be using here today.
So as we know the Billionaire Big Tech CEOs want their AI in
EVERYTHING, and Google apparently won a contract with the Pentagon to
put a version of Google Gemini AI in front of several million U.S.
armed forces members and other associated workers. This is being
called "genai.mil" and I'm told it's designated for CUI - Controlled
Unclassified Information and IL5 - Impact Level 5.
Secretary Hegseth and Google really made dramatic announcements about
this and the Pentagon apparently has AI-generated posters of Hegseth
in an Uncle Sam "I want you" pose saying "I want you to use AI!" and
there are emails and popups and other promotions pushing use of this
Google AI system for the military.
The announcements did raise many eyebrows because, among other
reasons, Hegseth and Google didn't seem to quite agree on what this
project is actually for. Hegseth talked in terms of increasing
lethality, and Google seemed to be saying it's really for
administrative purposes and some analyses -- but really oriented more
toward routine paperwork and research.
Stories did come out quickly claiming that persons with access to the
Google Gemini Military System asked it about the legality of that
controversial "double tap" boat strike, and Gemini said it was
illegal.
Now whether or not that's correct this does quickly take us to the
heart of the matter which is as we all know how often Gemini and other
AI systems are completely or partly wrong in their answers -- the
latter being even more dangerous.
AI disclaimers often admit that you can't trust AI to be accurate, so
it's understandable that there are concerns about bringing what's
often called AI Slop into a military context, even if restricted to
administrative operations.
I had an incident with Google's AI yesterday, not a critical one but
something to ponder. I asked Google when an organization I know was
taking their holiday break, usually some weeks. And the Google AI
Overview came back and told me their break was from December 22nd to
January 5th, and added that the duration was three weeks. And I'm
looking at this and thinking THREE weeks? Yeah, Google AI in this case
didn't know that 22 December to 5 January was only two weeks. Now if a
student had gotten something like this wrong they'd probably get an F.
But you do have to wonder how Google AI could make an error like that.
It's just a simple calendar query, no external information or context
required. If you can't depend on AI for accuracy in something so
simple and obvious, it's completely reasonable to be concerned about
more complex information where errors would NOT be so obvious.
Thinking about the impact of AI loaded with misinformation in military
situations can be rather chilling to say the least. While this isn't a
sci-fi movie style evil AI, given what we all DO know about REAL AI
now, there seems to be an almost endless array of ways that this
Google/Pentagon partnership to push military AI could go very, very
wrong indeed. And THAT'S a very serious concern.
- - -
L
- - -
--Lauren--
Lauren Weinstein
lauren@vortex.com (https://www.vortex.com/lauren)
Lauren's Blog: https://lauren.vortex.com
Mastodon: https://mastodon.laurenweinstein.org/@lauren
Signal: By request on need to know basis
Founder: Network Neutrality Squad: https://www.nnsquad.org
PRIVACY Forum: https://www.vortex.com/privacy-info
Co-Founder: People For Internet Responsibility
_______________________________________________
privacy mailing list
https://lists.vortex.com/mailman/listinfo/privacy
