NSA spending billions annually to test AI models, lawmakers told
Classified estimates reveal the agency's AI testing costs far exceed prior regulatory proposals, stoking debate over who should pay for oversight.
What to know
- NSA is spending billions annually to test frontier AI models for national security vulnerabilities — far exceeding estimates for comprehensive federal AI regulatory systems, which had been pegged at $20–36 million.
- The Trump administration is resisting new federal AI oversight, even as NSA testing costs mount and computing demands soar.
- Debate is intensifying over whether AI companies should fund their own regulatory review versus taxpayers bearing the cost; Anthropic and OpenAI have expressed openness to greater oversight, while Musk advocates for industry self-regulation.
Lack of transparency undermines accountability for billions in taxpayer spending on classified AI testing.
-
“the NSA is spending billions on testing AI models. But the details are classified. We don't know what for, which models, safety standards, nothing”
@timkellogg.me · Bluesky ↗
“For security reasons, the Department does not discuss the technical architecture or resource allocation for its AI tools”
Defense Department spokesperson, Official statement · The Washington Sun ↗
National Security Agency AI testing and oversight agencyTrump administration Federal executiveAnthropic Frontier AI labOpenAI Frontier AI lab
Elon Musk Tech leader
How it unfolded 1 development · click the chart to see its coverage articlesposts
-
1
Debate intensifies over whether AI companies should fund their own regulatory review
The high costs are driving renewed discussion about whether frontier labs should bear the financial burden of safety review rather than taxpayers. Some experts propose a government tax on AI companies to fund independent safety operations. Anthropic and OpenAI have indicated openness to greater federal oversight and potential industry funding, while Elon Musk and others advocate for industry self-review without government involvement.
“It also may intensify calls to have the frontier labs to shoulder the financial burden of reviewing the models, especially with spending set to only balloon as the technology grows more advanced…”
— The Washington Sun -
The 4th Amendment in practice, as really any law, cannot preemptively stop the bad practices of actors working in bad faith. It's remediative, as in, the evidence it collects at least in theory cannot be admitted as evidence in court. If you look at indictments, informations, any initial charging document, you'll see at least some if not a deluge…
2 more of the top 3 · 17 posts in this stretch
-
H
Millions estimated: NSA spends billions of US dollars on AI testing The NSA is testing the best AI models and will pay billions of US dollars for it in 2026. The US Congress had assumed this would cost a few million. https://www. heise.de/en/news/Millions-esti…
-
Scoop: The National Security Agency is spending billions this year on testing AI models - far more than previously known Per a classified NSA estimate described by sources Compute costs for AI have exploded - including for taxpayers https://www.washingtonsun.com/ ...
-
-
background
Trump administration resists federal AI oversight while NSA expands testing — President Trump has rejected regulatory proposals from Congress and frontier labs to impose new safeguards on AI models. The NSA's Artificial Intelligence Security Center began expanded testing following high-profile hacking and security incidents.
-
background
NSA costs dwarf prior regulatory proposals, raising stakes for oversight debate — The disclosed spending is significantly greater than previously known and has prompted lawmakers to estimate that comprehensive federal AI regulatory systems could cost tens of billions per year. Congressional Budget Office estimates for prior AI oversight bills ranged from $20 million to $36 million over five years — orders of magnitude lower than NSA's current annual testing expenditures.
-
background
NSA reveals billions in annual AI testing spending to lawmakers — The National Security Agency informed lawmakers it is spending billions of taxpayer dollars this year evaluating and testing advanced AI models to identify national security vulnerabilities, according to two sources familiar with classified intelligence estimates. The exact dollar amount remains undisclosed for security reasons.
Also covered reported alongside — the timeline has no entry for these yet
-
first by The Washington Sun, 2d ago · also HN Frontpage
1 more headline
- Classified Estimates Show the NSA Is Paying Billions to Test AI Models HN Frontpage · 1d ago
and 1 smaller piece
What people are saying 14 voices from 3 sites · best of 17 · verbatim
- Yesterday
-
I seem to recall a certain former NSA affiliated person speaking on the conjecture of NSA Full Capture from fiber optic taps at a HOPE conference many many years ago. The statements alluded to automated custom robotic “drive shufflers” for even the tape and drive libraries already in use at that time, and heavily implied that in some cases exotic…
-
What do they test? In general, I don't understand the whole "AI safety" stuff. What are the scenarios we need to be protected from and how is this going to be accomplished?
-
Previously it was a little reassuring to understand that the Eye of Sauron required some human analyst to actually look at your data.
-
An underappreciated aspect of LLMs is how good they are at zero-shot classification. I'd expect this is extremely good at monitoring communications streams, particularly in foreign languages.
-
Because they are watching everything, or most of it anyway, it would be really funny to know what kind of stuff they *actually* care about... and all the massive long string of crap that is considered incredibly abhorrent but just doesn't make the mark...
-
Obviously. The amount of data they hold on the American public alone is massive. They can't possibly read it all, they need a system that can pull out "dangerous" people, and find dirt on whoever they want to target, and AI will happily do that, badly. I just hope they actually take the time to review what it gives them for accuracy so that they…
-
The "it only counts as a search if we find something" reasoning out of the FISA courts is batshit insane, dodging the 4th Amendment like bullets in The Matrix.
-
Does anyone here happen to know the ballpark number of how much the three letters agencies are currently paying OpenAI/antropic/who ever per year/month..?
-
I guess I would be more surprised if the headline said "turns out NSA dropped the ball on AI and haven't even fired up a single LLM instance yet".It's mostly about "testing" and "billions". But I am sure "testing" here is in quotes just like I wrote it. They are not going to get an itemized list unless it's coming from Snowden 2.0- We "tested AI"…
-
I feel like any AI regulation passed wouldn't have enough teeth to rein in the NSA because not even the Fourth Amendment seems applicable to the agency. let's not forget the ex-post-facto legalization of mass wiretapping and surveillance [0] with both establishment political parties supporting it on 'national security' concerns [1]. whatever…
-
"Testing"? Sure, Jan.They would have to be incompetent if they were not using them both for security penetration as well as social engineering / election meddling as everyone else is.
- Sep 24
-
T
Sources: the NSA told lawmakers it is spending billions this year to test AI models; proposals for a US AI regulatory body estimated costs of $20M-$40M per year (Jeff Stein/The Washington Sun) https://www. washingtonsun.com/technology/c lassified-estimates-nsa-paying-billions-to-test-ai-models http://www. techmeme.com/260924/p37#a26092 4p37
-
the NSA is spending billions on testing AI models. But the details are classified. We don't know what for, which models, safety standards, nothing — www.washingtonsun.com/technology/ c...
-
P
NSA Paying Billions to Test AI Models https:// politicalwire.com/2026/09/24/n sa-paying-billions-to-test-ai-models/?ap=m