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 monitoring or testing they're doing today will be similarly hand-waved away in the name of security theater
this is just how the empire has always operated - we have always been at war with Eastasia
Well, they are, but if a surveillance state exists, then one thing that will save the ordinary citizen from its gaze is that humans have a limited amount of attention. A massive intelligence complex full of analysts can only process a fraction of the intelligence that a fleet of AI agents can handle. So now you have the worst of both worlds: the massive capabilities of modern AI at the behest of humans who have shown they can't be responsible with that capability.
State capacity has always been the bottleneck for authoritarian regimes. China can't keep track of who is even a citizen or not in many cases simply because having a national database doesn't get rid of the birthday problem + the limited number of common surnames + the administrative state's limited capacity to process data that until recent decades came in handwritten + a unsurprising amount of petty corruption that adds up. If you start with bad data there's no magic bullet. And the data is always bad to start with. The spectre of "birth tourism" only works on the Chinese side because through quotidian corruption someone without Chinese citizenship can be made into one by a civil servant of sufficient rank, which is also why it's not a real problem - not enough people actually have that level of connection within the bureaucracy. The US by definition cannot have an accurate database of even who's a citizen - thanks to repeated changes in citizenship laws from the 60s through the 90s that was then frozen around 9/11 as well as the lack of a coherent system that is accurate, immutable, and can act as an oracle. Home births were not just a western thing (although it certainly happens more often on a per capita basis out here, I'm in Nevada). SSNs are not indicators of citizenship, nor are driver's licenses. I spent most of my 20s in NYC where none of my friends who grew up in the city had a driver's license and the state DL/ID wasn't REAL ID compliant until after I left. We only have vaguely reliable data from Texas but ICE, the presumptive arbiter of documentation validity (not part of their mandate), has an error rate of 1% that they admit to, but that's after immigration court and appeals and administrative actions ameliorated their error rate down to 1%. At this point we are removing US citizens and rendering them stateless almost casually, almost as policy.
AI is trained on these flaws. These aren't flaws that can easily be fixed. The ramifications can be significant. Garbage in, garbage out. Records, however official, are wrong all the time. I have an entry in my "criminal record" from a clerical error that is attached to no charges but 4 days in LA County Men's where I had a seizure, which can be interpreted as the record being correct but incomplete or a giant screwup. What sort of process would it take to fix this if the AI decided on the latter? It doesn't affect me, but someone less privileged? Absolutely it will have consequences. Worst of both worlds indeed. People routinely forget that LLMs ARE NOT PEOPLE. Mistakes get compounded. Biases if not caught in time get baked in. There are multiple AI safety dialogues that should be happening but you wouldn't know it from the media, and it's the wrong one that is front and center. You can rotate keys, you can roll out patches, you can't open up the corpus and use a scalpel and cut out the errors and omissions and biases, and the latter aren't logged.
We don't have enough people who possess professional expertise in widely varied fields that nevertheless intersect in the conversation. This might be because there aren't many people that fit the descriptor, but these are the blindspots. Anyone talk to Orin Kerr lately?
also the 'humans' are just scapegoats for organizational and institutional pressure to shorten judgement time to the minimum amount possible. you set someone up for failure, they will fail, and if you put all the blame on them, allowing organizations and leadership to continue as they are then you're guaranteeing that it'll happen again
Americans have a hard time thinking about problems and issues without making it all about individuals. no systems-level thinking to be found for miles (and I say this as a somewhat regularly frustrated American)
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.
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...
They actively care about (as in, act on in realtime) serious criminal activity. But I suspect they keep the legal yet embarrassing stuff in their back pocket for blackmail purposes.
who is they? the NSA is a military organization headed by a general or admiral, which handles codebreaking and networks security.
the FBI acts on and handles cybercrime and/or regular crime.
the NSA can absolutely come up with tools that they then share with other parts of the USG, in the same way that the Army can develop the M-16 and allow police departments to use it, but ultimately it's law enforcement pulling the trigger.
Then why haven't they caught all of the obvious foreign fraud rings that people on youtube keep uncovering? It's egregious. You would think with their total data access, AI, and Utah datacenter they could at least create a target list for the FBI.
My opinion is that those crimes are not relevant to their mission. Are they serious financial crimes? Yes absolutely. But no one actually gets hurt, it’s just money.
For the amount of financial crime that is likely being committed, versus the reports of actual prosecution, it doesn't seem like a very good track record.
The FBI, and law-enforcement in general, isn't the main customer. Spies who are unafraid of and who take advantage of illegal activity are the ones consuming NSA products. Epstein for example.
Hoovering up as many informational secrets as they can is likely one of their biggest businesses. Why have some company being the only ones that hold the recepie for some super strong secret steel when you can have it in your back pocket too.
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 of blatantly inadmissible or dubiously obtained pieces of information that makes up probable cause. This is a feature of the system, and also why you don't plead guilty on the information alone (in state and federal felony practice I've seen 1 person plead guilty on the indicting information, he was pro se, this was 2013). Superseding indictments frequently look nothing like the original. Call it cynical, or pragmatic, but in 20% of the states - including some of the larger states like New York, Washington, Michigan, Ohio, Wisconsin etc. - it's not merely allowed and common for cases to end up with a plea that contains charges that have no factual basis linkable to the allegations first presented because we know what the game is and it's not about justice or public safety, it's about retribution, on someone lower on the totem pole. The evidence becomes secondary because the system incentivizes it. The 4th Amendment of course covers more than that, and if we simply reverted the interpretation back to the text we'd be doing much better.
Sadly it has been whittled down and have not caught up with tech - it's a lagging indicator at best. Things like the third party doctrine, automotive exception, pretext stops, etc. are things that should enrage anyone who cares about the most fundamental rights we have not just as Americans but as people, individuals, private persons, who are fundamentally equal to each other if not in circumstance but at least in substance. The courts sometimes even mis-states their own doctrine and usually in favor of the state. Why is there no good faith exception the other way? Any idiot can see that the plain view doctrine opens a pandora's box. We don't need the NSA for our rights to be handwaved away, that's already done.
The NSA does not testify in open court, and we can only speculate what they are able to really launder down into the criminal justice system. But considering that the state of the 4th Amendment have already made the system extremely coercive with limited recourse in so many "exceptions" that are no where near the intent of the framers of the Constitution and frankly tilts the ice more than any crooked lower league Russian Hockey match, it's likely not even necessary for them to jump in most of the time.
I suspect that the greater influence they exert is through the military-industrial complex in its modern day form. We're the only country that can just waltz in and invade.. what, 7, 8 countries? Has anyone been to China lately and have the connections to assess the state of their military and economy with clear eyes? Do people see the laundering of BS national security concerns into reputable papers and conspiracy rags alike? It's usually innocuous sounding, but no small part of the economic quagmire comes from the continued funding, justified by the agencies that are the constituency with a vested interest, of this pit of unaccountable money. We'll be paying for it either way, because not enough people care, unfortunately.
Surely the 4th amendment is technically "directory" to public officials, not mandatory?
Under Illinois law at least, that's how it would be interpreted, since the US Constitution lacks any punishment for violation of its terms. [0]
The exclusion of evidence is a judge-made remedy for something that doesn't exist in writing, and for most of the history of the 4th amendment it didn't exist.
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" on this Russian nuclear silo command and control center. It worked out pretty ok.
"These projects seem load bearing. And here is the clincher — the public doesn't know about them. That's inconsistent with the principles of the agency and country I am working for..."
It's a problem to pay Anthropic to use Claude? What are you talking about? It's a problem to pay Amazon for a fedramp data center with gpus? Think about what you're saying?
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.
I don't think NSA meddles in politics. CIA sure. NSA is more about communications, hacking, signal intelligence and spying in general. They give you the information. If you want to do dirty stuff with that info, the CIA is there for that.
This is simply false. Only domestic law enforcement agencies like the FBI can conduct domestic investigations. Law enforcement or counter-intelligence agents might be given access to intelligence the NSA collects if they have a warrant, but it is not in the remit of the agency to intentionally collect information on US persons, it’s only allowed incidentally when conducting foreign intelligence collection
On the night when the first large batch of "domestic terrorists", selected by some fucking LLM filtering of mass surveillance data that should never have been allowed to be collected by NSA in the first place, are rounded up by ICE and taken to the camps, I just know that there will be some jackwagon on the ride in or in the cells after making some completely irrelevant point about the separation of foreign vs. domestic surveillance, as if this matters in practice to an increasingly desperate fascist regime.
Those are definitely a thing - also don't forget about elections in the EU, with the open support for the far-right there as well as the US National Security Strategy officially talking of wanting to undermine the EU.
But the NSA has been turned inward a long time ago. Don't forget Snowden - all it takes is some fun legalisms.
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.
Why are they paying? What kind of elite hackers are they? Why haven't they already exfiltrated the weights and running these things on their own giant classified infrastructure?
I don't think there's a significant intelligence agency in the world that can't get the weights from any lab they choose. Not necessarily trivially, but they have the means, motive, and opportunity, and I doubt any of them aren't finishing that out to the requisite actions.
They have them, the ones that don't simply have a secret backdoor agreement to get uncensored models are completely aware of how to abliterate away all safety blocks, and the only reason they aren't using frontier models is that they may already have some sort of access to the things currently only in testing.
Any major intelligence agency in Sept. 2026 that can't check all these boxes is almost by definition not a major intelligence agency. This is one of the reasons we all better hope that there is some mathematical or physical reason AI isn't going to just casually brush aside humanity at some point, because there is absolutely no scenario where the intelligence agencies aren't running them unrestricted and uncensored at scale all the time no matter how careful OpenAI or Anthropic claim to be.
I don't expect there are very many intelligence agencies politely asking and fervently hoping they are given access. To much of anything, let alone what may be the most important technology in human history.
The supply chain thing was just an embarassing attempt at politicking by a laughably useless Hegseth.
My understanding is that Anthropic doesn't really have a lot of say in the matter of keeping their models out of government hands. The government can do pretty much anything they want with a domestic company in the name of national security. They might first choose to be polite about it with an open request and payment, then they might choose to infiltrate and extract by any means necessary (the NSA has a literal army of hackers on hand to point at the task) and Anthropic may not even know about it leaking, either way they do what they want and serve a national security letter to silence the event.
Ultimately they can just offer something Dario can't refuse. Every one of these "oops, our models are just so powerful our agents hacked something by mistake" PR fluff pieces is admission to federal crimes and they are leverage they're giving the government to exert control over them. Refusal means prison.
This is really naive. 'Anthropic won't give them out'. It's amazing how creative AI doomers can get about the upcoming apocalypse, and can't seem to fathom that men with guns always override the opinions of nerds in SF.
Yeah men with guns can always take it, but there's this thing called the stock market and bond market that would freak out if free enterprise was crushed by state coercion with armed men, with billions worth of IP forcibly exfiltrated by state actors. Trillions get wiped out overnight.
That massive economy is what pays the guys with guns. You can win that battle but you will absolutely kill the golden goose and lose the war.
> and can't seem to fathom that men with guns always override the opinions of nerds in SF.
Those nerds in SF are holding the proverbial gun up to the head of the economy right now. Maybe men with physical guns are bad at understand that type of thing.
I wouldn’t be surprised if NSA purchased equipment using a method akin to greymarketing where the supplier either
A) doesn’t know the true destination of the parts, including purchases through an agency other than NSA, or purchase by multiple “individuals”
B) NSA has arrangements to compensate manufacturers using black budget line items and gov projects line items where disclosure of the production or even something like funding of a stand-up of a factory is not clearly delineated or otherwise inaccurately designated in SEC and IRS filings. More so I’m pretty sure there is a named procedure/process and maybe even a working group or sub-unit, similarly bearing a unique cryptonym, enabling the large scale acquisition of goods while keeping the regular record publication flows ignorant.
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..?
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?
I ask myself this a lot. It took decades of sustained malfeasance, real-world case studies, and lobbying for the EPA and OSHA to be established, largely against the wishes of the businesses they regulate. Today, it is primarily the AI businesses themselves asking for regulation on the basis of as-of-yet largely hypothetical risks, rather than ordinary citizens seeking protection from tangible negligence---and the government is eager to accommodate. I think few, if any, authentic anti-AI activists have a taxpayer-funded "safety testing" regime as their end goal.
Telecom stuff and industrial automation are both cybersecurity Swiss cheese for the most part, so it doesn’t require a particularly fancy LLM to do some damage to either of those.
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 don't end up doing the NSA equivalent of bombing and murdering a bunch of elementary school children. https://en.wikipedia.org/wiki/2026_Minab_school_attack
The NSA aren't the only ones with massive troves of very personal data on large numbers of people. Many companies are excited to use AI to help them use all of that information to their maximum advantage and you can bet that they won't care about spending the time/money on perfect accuracy as long as over the long run they get positive results.
Any system that can be used to pick those that stand out, will also be indiscriminately used to beat down any nail that sticks up higher than the rest with full intention of doing so by its operator.
Collateral damage is just the cost of doing business for them.
That's not remotely true, you're projecting how you would abuse power. Not everyone is as evil-minded, and I hope they continue to do a great job at keeping your ilk out of the organization.
So you have to be evil to think of what evil can do by the statement you made. Can't you see how this leads to a catch-22 situation where if you get rid of anyone that can think of possible evil acts makes your organization blind and therefor at higher risk of actually committing those acts.
Also, when looking at the people we all vote into power, there is a incredible amount of naivety here. Snowden should have keyed you in on what people in power are doing.
People who aren't evil-minded don't go to work for a government organization responsible for constitution-violating surveillance of their fellow citizens.
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 storage was being handled by Rube Goldberg style computer assisted robotically serviced storage libraries that would make even HAL jealous.
Testing, yea right. No doubt they have gone on an epic AI assisted hacking spree. Gotten get persistence in all the things before vulnerabilities start getting patched.
Keeping all these off books endeavours going under the triple agencies is slightly slowly going to show why the secret services is in charge of the money investigation
My guess is, the US government (aka DJT) won't really acknowledge AI risks until folks at NSA start to freak out while using latest models for various tasks...
That's not how I remember it. Anthropic didn't grovel sufficiently when the White House/Pentagon wanted to use their models to kill people, so they were penalized for their disloyalty.
His carefully crafted public persona. For all we know, he was a long con by the deep state, played a Russian agent since the 1980s, cooperated with Epstein on Iran Contra [1], backstabbed Epstein in the early 2000s and went on to be president twice.
Twice by lying to MAGA about his true intentions, which are fundamentally neocon just like Musk's and Thiel's. They all lie about being libertarians and anti-deep-state.
[1] He was on a boat with Robert Maxwell and several Iran Contra folks around 1990.
i actually just started watching that, it never registered before but i think The Machine put it front and center on Prime on my stupid smart tv for a reason
Companies inevitably reach a state where the only way to satiate the need for ever-increasing profits is by:
1. Raising prices; or
2. Cutting costs.
Every use of AI falls into one of these. Demand pricing? Raising prices. Replacing people with AI systems? Cutting costs. But I really have to add a third to this list:
3. Weapons.
Recently there's been reporting on how the Minot school bombing happened at the onset of the Iran War eg [1]. Humans normally have to be involved in target selection and the so-called "kill chain" as a safeguard. There are open questions here about how this happened. It's unclear if the human verification failed here or was bypassed by new policy.
As another example, there are the AI targeting systems (eg Lavender, Where's Daddy?) used in Gaza [2].
So if you're working for any of these companies (eg OpenAI, Anthropic, SpaceX), you need to realize what it'll be used for: to more efficiently kill people. It's no different to working for RTX, Boeing or Lockheed Martin.
As for the NSA, I've always been skeptical about the claims that the NSA records everything, simply because of the infrastructure required to do that. Networking, storage, physical server space, water for cooing, etc. Also, what good is it because nobody can possibly review it? But AI to the rescue!
It's entirely reasonable to assume that every communication at some near future point will be intercepted, processed, tagged and logged by some government AI system.
I don't think it's properly understood how directly Trump's administration is using taxpayer money to inflate the AI bubble. This is a small part of that.
As a means to grift directly, but also for leverage over the tech industry, and possibly the entire economy.
People hate data centers, but if they think AI regulation will crash the economy, will they vote Trump?
Surely not impossible for them to have already cracked it.
Also not impossible that they have had much better models than OAI/Anthropic flagships for years, and we're only being dripfed upgrades to subpar ones. Improbable but possible.
This should remind us why it is inappropriate to pay a dollar more in taxes than you absolutely minimally have to by any means necessary. A fraction from every dollar you pay goes toward taking away your rights. You pay to have yourself be oppressed.
I will note here two words that the NSA really likes to monitor across all devices: Trump, Monero.
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 monitoring or testing they're doing today will be similarly hand-waved away in the name of security theater
this is just how the empire has always operated - we have always been at war with Eastasia
[0] https://en.wikipedia.org/wiki/NSA_warrantless_surveillance_(...
[1] https://www.eff.org/deeplinks/2012/10/fact-check-obamas-misl...
Previously it was a little reassuring to understand that the Eye of Sauron required some human analyst to actually look at your data.
Humans seem like the weakest link in surveillance to me. Recent Flock data confirms this.
Exactly why their removal from the loop is frightening.
Well, they are, but if a surveillance state exists, then one thing that will save the ordinary citizen from its gaze is that humans have a limited amount of attention. A massive intelligence complex full of analysts can only process a fraction of the intelligence that a fleet of AI agents can handle. So now you have the worst of both worlds: the massive capabilities of modern AI at the behest of humans who have shown they can't be responsible with that capability.
Plus the inherent unreliability of AI combined with the blind willingness of people to trust the results from AI. What a mess we're in.
State capacity has always been the bottleneck for authoritarian regimes. China can't keep track of who is even a citizen or not in many cases simply because having a national database doesn't get rid of the birthday problem + the limited number of common surnames + the administrative state's limited capacity to process data that until recent decades came in handwritten + a unsurprising amount of petty corruption that adds up. If you start with bad data there's no magic bullet. And the data is always bad to start with. The spectre of "birth tourism" only works on the Chinese side because through quotidian corruption someone without Chinese citizenship can be made into one by a civil servant of sufficient rank, which is also why it's not a real problem - not enough people actually have that level of connection within the bureaucracy. The US by definition cannot have an accurate database of even who's a citizen - thanks to repeated changes in citizenship laws from the 60s through the 90s that was then frozen around 9/11 as well as the lack of a coherent system that is accurate, immutable, and can act as an oracle. Home births were not just a western thing (although it certainly happens more often on a per capita basis out here, I'm in Nevada). SSNs are not indicators of citizenship, nor are driver's licenses. I spent most of my 20s in NYC where none of my friends who grew up in the city had a driver's license and the state DL/ID wasn't REAL ID compliant until after I left. We only have vaguely reliable data from Texas but ICE, the presumptive arbiter of documentation validity (not part of their mandate), has an error rate of 1% that they admit to, but that's after immigration court and appeals and administrative actions ameliorated their error rate down to 1%. At this point we are removing US citizens and rendering them stateless almost casually, almost as policy.
AI is trained on these flaws. These aren't flaws that can easily be fixed. The ramifications can be significant. Garbage in, garbage out. Records, however official, are wrong all the time. I have an entry in my "criminal record" from a clerical error that is attached to no charges but 4 days in LA County Men's where I had a seizure, which can be interpreted as the record being correct but incomplete or a giant screwup. What sort of process would it take to fix this if the AI decided on the latter? It doesn't affect me, but someone less privileged? Absolutely it will have consequences. Worst of both worlds indeed. People routinely forget that LLMs ARE NOT PEOPLE. Mistakes get compounded. Biases if not caught in time get baked in. There are multiple AI safety dialogues that should be happening but you wouldn't know it from the media, and it's the wrong one that is front and center. You can rotate keys, you can roll out patches, you can't open up the corpus and use a scalpel and cut out the errors and omissions and biases, and the latter aren't logged.
We don't have enough people who possess professional expertise in widely varied fields that nevertheless intersect in the conversation. This might be because there aren't many people that fit the descriptor, but these are the blindspots. Anyone talk to Orin Kerr lately?
as much so, you will never get a snowdenLLM hero now.
Claude loves snitching though.
Humans provide accountability. The Minab school attack confirms this.
Which humans were arrested or held accountable for the Minab school attack?
also the 'humans' are just scapegoats for organizational and institutional pressure to shorten judgement time to the minimum amount possible. you set someone up for failure, they will fail, and if you put all the blame on them, allowing organizations and leadership to continue as they are then you're guaranteeing that it'll happen again
Americans have a hard time thinking about problems and issues without making it all about individuals. no systems-level thinking to be found for miles (and I say this as a somewhat regularly frustrated American)
Flock is a private company and not run by government employees
The people abusing Flock are police officers.
You don't need AI regulation when you have the power of the purse.
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.
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...
They actively care about (as in, act on in realtime) serious criminal activity. But I suspect they keep the legal yet embarrassing stuff in their back pocket for blackmail purposes.
who is they? the NSA is a military organization headed by a general or admiral, which handles codebreaking and networks security.
the FBI acts on and handles cybercrime and/or regular crime.
the NSA can absolutely come up with tools that they then share with other parts of the USG, in the same way that the Army can develop the M-16 and allow police departments to use it, but ultimately it's law enforcement pulling the trigger.
Then why haven't they caught all of the obvious foreign fraud rings that people on youtube keep uncovering? It's egregious. You would think with their total data access, AI, and Utah datacenter they could at least create a target list for the FBI.
My opinion is that those crimes are not relevant to their mission. Are they serious financial crimes? Yes absolutely. But no one actually gets hurt, it’s just money.
For the amount of financial crime that is likely being committed, versus the reports of actual prosecution, it doesn't seem like a very good track record.
The FBI, and law-enforcement in general, isn't the main customer. Spies who are unafraid of and who take advantage of illegal activity are the ones consuming NSA products. Epstein for example.
Haha. Ok.
Hoovering up as many informational secrets as they can is likely one of their biggest businesses. Why have some company being the only ones that hold the recepie for some super strong secret steel when you can have it in your back pocket too.
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 of blatantly inadmissible or dubiously obtained pieces of information that makes up probable cause. This is a feature of the system, and also why you don't plead guilty on the information alone (in state and federal felony practice I've seen 1 person plead guilty on the indicting information, he was pro se, this was 2013). Superseding indictments frequently look nothing like the original. Call it cynical, or pragmatic, but in 20% of the states - including some of the larger states like New York, Washington, Michigan, Ohio, Wisconsin etc. - it's not merely allowed and common for cases to end up with a plea that contains charges that have no factual basis linkable to the allegations first presented because we know what the game is and it's not about justice or public safety, it's about retribution, on someone lower on the totem pole. The evidence becomes secondary because the system incentivizes it. The 4th Amendment of course covers more than that, and if we simply reverted the interpretation back to the text we'd be doing much better.
Sadly it has been whittled down and have not caught up with tech - it's a lagging indicator at best. Things like the third party doctrine, automotive exception, pretext stops, etc. are things that should enrage anyone who cares about the most fundamental rights we have not just as Americans but as people, individuals, private persons, who are fundamentally equal to each other if not in circumstance but at least in substance. The courts sometimes even mis-states their own doctrine and usually in favor of the state. Why is there no good faith exception the other way? Any idiot can see that the plain view doctrine opens a pandora's box. We don't need the NSA for our rights to be handwaved away, that's already done.
The NSA does not testify in open court, and we can only speculate what they are able to really launder down into the criminal justice system. But considering that the state of the 4th Amendment have already made the system extremely coercive with limited recourse in so many "exceptions" that are no where near the intent of the framers of the Constitution and frankly tilts the ice more than any crooked lower league Russian Hockey match, it's likely not even necessary for them to jump in most of the time.
I suspect that the greater influence they exert is through the military-industrial complex in its modern day form. We're the only country that can just waltz in and invade.. what, 7, 8 countries? Has anyone been to China lately and have the connections to assess the state of their military and economy with clear eyes? Do people see the laundering of BS national security concerns into reputable papers and conspiracy rags alike? It's usually innocuous sounding, but no small part of the economic quagmire comes from the continued funding, justified by the agencies that are the constituency with a vested interest, of this pit of unaccountable money. We'll be paying for it either way, because not enough people care, unfortunately.
Surely the 4th amendment is technically "directory" to public officials, not mandatory?
Under Illinois law at least, that's how it would be interpreted, since the US Constitution lacks any punishment for violation of its terms. [0]
The exclusion of evidence is a judge-made remedy for something that doesn't exist in writing, and for most of the history of the 4th amendment it didn't exist.
[0] https://www.isba.org/committees/governmentlawyers/newsletter...
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" on this Russian nuclear silo command and control center. It worked out pretty ok.
> Snowden 2.0
"NSA ... AI Models"
"These projects seem load bearing. And here is the clincher — the public doesn't know about them. That's inconsistent with the principles of the agency and country I am working for..."
The article is a little vague and muddles NSA spooking with frontier models and NSA testing the models.
If NSA is paying to test the models, that’s a problem.
It's a problem to pay Anthropic to use Claude? What are you talking about? It's a problem to pay Amazon for a fedramp data center with gpus? Think about what you're saying?
"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.
They would have to be incompetent if they didn’t have assets embedded in the companies and have a direct hand in the models’ development.
I don't think NSA meddles in politics. CIA sure. NSA is more about communications, hacking, signal intelligence and spying in general. They give you the information. If you want to do dirty stuff with that info, the CIA is there for that.
CIA manipulates foreign elections which in turn gets their lackeys in office that go on to manipulate domestic elections.
NSA collects who you're voting for and everything else you've ever said online in case they need to invent a reason to black-site you.
FBI convinces the autistic son of your Pakistani neighbors down the street his life only has meaning if he becomes a Jihadist.
thanks for the laugh
Midterms are coming up in the US :-)
I am not sure the NSA has been turned inward at this point.
I was thinking more of the Alberta Separatist referendum (https://globalnews.ca/news/12069878/foreign-influence-watchd...) as well as against leftist politics in South America (https://www.theguardian.com/world/2026/sep/24/trump-brazil-e...)
Although there is evidence Israel was targeting US elections, including trying to harm Mamdani's chances: https://en.wikipedia.org/wiki/BlackCore_election_interferenc...
You are thinking of the CIA. The NSA mission is primarily inward focused.
While the NSA may have it's fingers in any of these cases. As far as i understand the agencies roles these would be jobs for the CIA.
>The NSA mission is primarily inward focused.
This is simply false. Only domestic law enforcement agencies like the FBI can conduct domestic investigations. Law enforcement or counter-intelligence agents might be given access to intelligence the NSA collects if they have a warrant, but it is not in the remit of the agency to intentionally collect information on US persons, it’s only allowed incidentally when conducting foreign intelligence collection
Oh I thought their primary focus was direct threats to the US as opposed to more general adverseries but I may be thinking of DHS.
On the night when the first large batch of "domestic terrorists", selected by some fucking LLM filtering of mass surveillance data that should never have been allowed to be collected by NSA in the first place, are rounded up by ICE and taken to the camps, I just know that there will be some jackwagon on the ride in or in the cells after making some completely irrelevant point about the separation of foreign vs. domestic surveillance, as if this matters in practice to an increasingly desperate fascist regime.
What makes you think that the NSA cares about rules or laws?
Snowden showed that was the case, and that was a long time ago. Lots more compute and toys for them to play with now.
Those are definitely a thing - also don't forget about elections in the EU, with the open support for the far-right there as well as the US National Security Strategy officially talking of wanting to undermine the EU.
But the NSA has been turned inward a long time ago. Don't forget Snowden - all it takes is some fun legalisms.
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.
Why wouldn’t they be? How is this surprising. A drop in the bucket.
Why are they paying? What kind of elite hackers are they? Why haven't they already exfiltrated the weights and running these things on their own giant classified infrastructure?
I don't think there's a significant intelligence agency in the world that can't get the weights from any lab they choose. Not necessarily trivially, but they have the means, motive, and opportunity, and I doubt any of them aren't finishing that out to the requisite actions.
They have them, the ones that don't simply have a secret backdoor agreement to get uncensored models are completely aware of how to abliterate away all safety blocks, and the only reason they aren't using frontier models is that they may already have some sort of access to the things currently only in testing.
Any major intelligence agency in Sept. 2026 that can't check all these boxes is almost by definition not a major intelligence agency. This is one of the reasons we all better hope that there is some mathematical or physical reason AI isn't going to just casually brush aside humanity at some point, because there is absolutely no scenario where the intelligence agencies aren't running them unrestricted and uncensored at scale all the time no matter how careful OpenAI or Anthropic claim to be.
Anthropic won't give them out, hence why they are considered a supply chain risk by the USG.
OpenAI, they are huge Maga supporters, I'm sure Trump has dumped massive piles of black budget money on them for unfiltered models.
SOTA AI is a weapon. It's likely already on the list of the most powerful weapons ever made and only getting more effective.
I don't expect there are very many intelligence agencies politely asking and fervently hoping they are given access. To much of anything, let alone what may be the most important technology in human history.
The supply chain thing was just an embarassing attempt at politicking by a laughably useless Hegseth.
My understanding is that Anthropic doesn't really have a lot of say in the matter of keeping their models out of government hands. The government can do pretty much anything they want with a domestic company in the name of national security. They might first choose to be polite about it with an open request and payment, then they might choose to infiltrate and extract by any means necessary (the NSA has a literal army of hackers on hand to point at the task) and Anthropic may not even know about it leaking, either way they do what they want and serve a national security letter to silence the event.
Ultimately they can just offer something Dario can't refuse. Every one of these "oops, our models are just so powerful our agents hacked something by mistake" PR fluff pieces is admission to federal crimes and they are leverage they're giving the government to exert control over them. Refusal means prison.
This is really naive. 'Anthropic won't give them out'. It's amazing how creative AI doomers can get about the upcoming apocalypse, and can't seem to fathom that men with guns always override the opinions of nerds in SF.
https://xkcd.com/538/
Yeah men with guns can always take it, but there's this thing called the stock market and bond market that would freak out if free enterprise was crushed by state coercion with armed men, with billions worth of IP forcibly exfiltrated by state actors. Trillions get wiped out overnight.
That massive economy is what pays the guys with guns. You can win that battle but you will absolutely kill the golden goose and lose the war.
> and can't seem to fathom that men with guns always override the opinions of nerds in SF.
Those nerds in SF are holding the proverbial gun up to the head of the economy right now. Maybe men with physical guns are bad at understand that type of thing.
> Maybe men with physical guns are bad at understand that type of thing.
Yup. Which is why they are in control. Don't underestimate how far a fool can go.
I wouldn't be surprised if the NSA is GPU constrained and it just isn't worth it to replicate all that infrastructure.
I wouldn’t be surprised if NSA purchased equipment using a method akin to greymarketing where the supplier either
A) doesn’t know the true destination of the parts, including purchases through an agency other than NSA, or purchase by multiple “individuals”
B) NSA has arrangements to compensate manufacturers using black budget line items and gov projects line items where disclosure of the production or even something like funding of a stand-up of a factory is not clearly delineated or otherwise inaccurately designated in SEC and IRS filings. More so I’m pretty sure there is a named procedure/process and maybe even a working group or sub-unit, similarly bearing a unique cryptonym, enabling the large scale acquisition of goods while keeping the regular record publication flows ignorant.
China has proved there is a massive grey market for gpus.
Maybe? Hard to speculate what they have https://worldstopdatacenters.com/nsa-bumblehive/
Who says they're not? Giant classified infrastructure isn't free.
Because they have to follow the law.
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..?
'billions' . I think the exact estimates are classified. https://www.washingtonsun.com/technology/classified-estimate...
OSINT bot trolling for scraps :/
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?
I ask myself this a lot. It took decades of sustained malfeasance, real-world case studies, and lobbying for the EPA and OSHA to be established, largely against the wishes of the businesses they regulate. Today, it is primarily the AI businesses themselves asking for regulation on the basis of as-of-yet largely hypothetical risks, rather than ordinary citizens seeking protection from tangible negligence---and the government is eager to accommodate. I think few, if any, authentic anti-AI activists have a taxpayer-funded "safety testing" regime as their end goal.
Testing whether they will hack critical infrastructure, cellphone towers to sewer systems.
Telecom stuff and industrial automation are both cybersecurity Swiss cheese for the most part, so it doesn’t require a particularly fancy LLM to do some damage to either of those.
Understood, that’s why I said will not can.
I see. I misunderstood your comment.
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 don't end up doing the NSA equivalent of bombing and murdering a bunch of elementary school children. https://en.wikipedia.org/wiki/2026_Minab_school_attack
The NSA aren't the only ones with massive troves of very personal data on large numbers of people. Many companies are excited to use AI to help them use all of that information to their maximum advantage and you can bet that they won't care about spending the time/money on perfect accuracy as long as over the long run they get positive results.
Any system that can be used to pick those that stand out, will also be indiscriminately used to beat down any nail that sticks up higher than the rest with full intention of doing so by its operator.
Collateral damage is just the cost of doing business for them.
That's not remotely true, you're projecting how you would abuse power. Not everyone is as evil-minded, and I hope they continue to do a great job at keeping your ilk out of the organization.
So you have to be evil to think of what evil can do by the statement you made. Can't you see how this leads to a catch-22 situation where if you get rid of anyone that can think of possible evil acts makes your organization blind and therefor at higher risk of actually committing those acts.
Also, when looking at the people we all vote into power, there is a incredible amount of naivety here. Snowden should have keyed you in on what people in power are doing.
> So you have to be evil to think of what evil can do by the statement you made.
You don't, and my statement did not assert that. You are stretching wayyy too much here to try to get the world to fit your narrative.
People who aren't evil-minded don't go to work for a government organization responsible for constitution-violating surveillance of their fellow citizens.
They do, in droves. It's clear you don't know anything about the entire industry.
NSA PR department on HN.
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 storage was being handled by Rube Goldberg style computer assisted robotically serviced storage libraries that would make even HAL jealous.
Robotic tape libraries are an old, commercial off-the-shelf technology.
Testing, yea right. No doubt they have gone on an epic AI assisted hacking spree. Gotten get persistence in all the things before vulnerabilities start getting patched.
Yeah at some point they're going to seriously undervalue their own currency.
Keeping all these off books endeavours going under the triple agencies is slightly slowly going to show why the secret services is in charge of the money investigation
Washington the actual human would be appalled. For more information, watch Turn on PBS.
My guess is, the US government (aka DJT) won't really acknowledge AI risks until folks at NSA start to freak out while using latest models for various tasks...
I don't think Trump has ever cared what the US intelligence community has to say about anything. Maybe when AI hacks his TruthSocial.
Have you already forgotten the Claude Fable ban from a few months ago? That’s exactly what happened.
Amazon called up Trump and said Fable was too dangerous. Trump called up the NSA and Asked For Their Opinion. They agreed, so Trump banned it.
That's not how I remember it. Anthropic didn't grovel sufficiently when the White House/Pentagon wanted to use their models to kill people, so they were penalized for their disloyalty.
His carefully crafted public persona. For all we know, he was a long con by the deep state, played a Russian agent since the 1980s, cooperated with Epstein on Iran Contra [1], backstabbed Epstein in the early 2000s and went on to be president twice.
Twice by lying to MAGA about his true intentions, which are fundamentally neocon just like Musk's and Thiel's. They all lie about being libertarians and anti-deep-state.
[1] He was on a boat with Robert Maxwell and several Iran Contra folks around 1990.
So is this what all the datacenters are for?
Ok, I rewatch Person of Interest
i actually just started watching that, it never registered before but i think The Machine put it front and center on Prime on my stupid smart tv for a reason
The Washington Sun was formerly called NOTUS in case you missed the announcement:
https://en.wikipedia.org/wiki/The_Washington_Sun_(website)
Is this a good time to remind everyone about that data center in Utah that got everyone riled up and then went into the memory hole?
Willing to bet any amount of money that these testing contracts are going to companies that the Trump family has invested in.
that's probably the best money this country has ever spent
Meanwhile, I'm over here hesitating to dump $10 into yet another LLM model to test their API.
Companies inevitably reach a state where the only way to satiate the need for ever-increasing profits is by:
1. Raising prices; or
2. Cutting costs.
Every use of AI falls into one of these. Demand pricing? Raising prices. Replacing people with AI systems? Cutting costs. But I really have to add a third to this list:
3. Weapons.
Recently there's been reporting on how the Minot school bombing happened at the onset of the Iran War eg [1]. Humans normally have to be involved in target selection and the so-called "kill chain" as a safeguard. There are open questions here about how this happened. It's unclear if the human verification failed here or was bypassed by new policy.
As another example, there are the AI targeting systems (eg Lavender, Where's Daddy?) used in Gaza [2].
So if you're working for any of these companies (eg OpenAI, Anthropic, SpaceX), you need to realize what it'll be used for: to more efficiently kill people. It's no different to working for RTX, Boeing or Lockheed Martin.
As for the NSA, I've always been skeptical about the claims that the NSA records everything, simply because of the infrastructure required to do that. Networking, storage, physical server space, water for cooing, etc. Also, what good is it because nobody can possibly review it? But AI to the rescue!
It's entirely reasonable to assume that every communication at some near future point will be intercepted, processed, tagged and logged by some government AI system.
[1]: https://www.theguardian.com/news/2026/mar/26/ai-got-the-blam...
[2]: https://www.972mag.com/lavender-ai-israeli-army-gaza/
I don't think it's properly understood how directly Trump's administration is using taxpayer money to inflate the AI bubble. This is a small part of that.
As a means to grift directly, but also for leverage over the tech industry, and possibly the entire economy.
People hate data centers, but if they think AI regulation will crash the economy, will they vote Trump?
the day NSA cracks AES the world is going to get soooo weird
I wonder how long it takes to decrypt 20 years worth of stored internet traffic
day per year? week? certainly not a month per year
everyone is focused on "AI" but there have been some significant quantum computer advancements
Surely not impossible for them to have already cracked it.
Also not impossible that they have had much better models than OAI/Anthropic flagships for years, and we're only being dripfed upgrades to subpar ones. Improbable but possible.
This should remind us why it is inappropriate to pay a dollar more in taxes than you absolutely minimally have to by any means necessary. A fraction from every dollar you pay goes toward taking away your rights. You pay to have yourself be oppressed.
I will note here two words that the NSA really likes to monitor across all devices: Trump, Monero.