I think the US intelligence community has had absolute unaccountable power to take key decisions of the US govt for the last 80 years (since world war 2), and I don’t see this changing by default.
Key decisions they unilaterally take include decisions on sanctions (on equipment, reagents, etc) to stay at the frontier on any technological race in last 80 years. Another key decision they unilaterally take includes decisions to start wars.
They are not accountable to the Congress / Senate, or to the Supreme Court, or to any regulatory body or similar, when it comes to these key decisions.
Heads of US intelligence already have active relationships with the heads of all major US AI companies. They are one of the primary reasons the US AI companies will remain unaccountable to the US public, unless something big changes.
I think social media influencers are most accountable to the public. Politicians and billionaires are less unaccountable than them. Heads of intelligence are the least accountable to the public. There are structural reasons why this unaccountability exists.
Persuading heads of US intelligence community to stop the race towards ASI is too hard, they are likely to accelerate towards ASI even if they have some understanding of the risks.
I have a fairly strong belief in this direction. Why?
I have studied every US govt whistleblower leaking classified info in the past 50 years as part of the work I published.
I have read atleast a couple books about who actually took the decisions during the Manhattan project.
I have read a bit about US cold war history, and a bit on more recent wars started by the US, to see who took the decisions.
I have also looked at some of the journalists, politicians and mass protest movements in the US trying to (unsuccessfully) hold the US intelligence community more accountable, in more recent times.
I have read the blogs of almost every famous cypherpunk activist, and have carefully tracked who got arrested versus not, and why.
I have gone through public statements of many recent heads of US intelligence on topics including AI, cybersecurity, social media, surveillance, drone warfare, and so on.
(Incomplete list) Examples where I think the US intelligence community was held even a little bit accountable to the public
I think the Vietnam war stopped sooner than it would have otherwise, because of mass protests.
My guess is that one factor for why Truman didn’t support von Nuemann’s plan of doing a pre-emptive first strike and creating a nuclear monopoly for the US, was that he was afraid of mass protests and the authoritarianism he would have to engage in to suppress these protests.
How can you change my mind here?
The best way to change my mind is obviously to actually get US IC members in support of an AI pause.
I have very little hope in the “inside game” strategy that EA billionaires are attempting, such as putting Jason Matheny in RAND or Paul Christiano in NIST or similar.
I have slightly more hope in being Loud and Annoying. Actually go organise protests at Fort Meade. Actually go figure out which social media all these US IC people use and persistently call them out there, for not taking a stance on AI pause.
The second best way to change my mind is to offer historical evidence like I have shared above. Although, given that I already have a large number of datapoints on my side, just one datapoint on the opposite side won’t be enough, you’ll need either many datapoints or a datapoint with a lot of weight.
Manhattan project—Read the wikipedia pages, read Making of atomic bomb by Richard Rhodes. Just go ask some historians in academia, I think what I am stating here is the consensus viewpoint, and is not that controversial or politicised.
US Cold War history—This stuff is a bit more politicised by the leftists (partly because counterculture movement took off in 60s). You can read Devil’s chessboard by David Talbot for left-leaning view on this.
Journalists, politicians, mass protest movements against the US IC today—I don’t have any good resources on this, but you’ll easily get plenty of (extremely politicised) sources on No Kings and anti-ICE protests and so on. Both left-leaning and right-leaning groups in US have organised various protests, there’s a long list of YouTubers and so on. Ask gpt-5.5 to compile it.
Cypherpunk activists—I probably do have the list of cypherpunk activists compiled somewhere in my notes, but I can’t find it at the moment. Again just go ask gpt-5.5, and actually go read the original blogs, not what the media wrote about them. Or go read the cypherpunks mailing list itself. This index has organised it by popular authors. The short version is that people working on encryption (PGP, Monero, Zcash, Tornado cash, NIST crypto standards) and leaking of classified information (Wikileaks, Blacknet, etc) are the ones who have been the most relentlessly targeted and imprisoned. If you’re just working on a secure OS or on blockchain more broadly or on dark web cloud infra more broadly, you’re less likely to be targeted. There is a long history here. For instance, civilian usage of encrypted radio was banned during or after world war 2, after both Axis and Allies used Numbers Stations to send messages, and it remains banned to this day. Digital encryption was initial regulated as a munition, but due to courage and effort of the PGP authors and a few other cypherpunk, PGP atleast exists now and is legal to use. Almost nobody uses it, and people working on such tech are at higher risk even today.
I am busy doing actual object-level work, and I have already wasted a year of my life arguing on lesswrong with random people with no skin in the game. One year is extremely costly given how close we are to the end of the world.
If someone who actually has expertise or power can commit some actual skin-in-the-game, I can spend more time talking to them. For example, if someone says they will pre-commit to devoting X hours of their life giving feedback, or someone will pre-commit to devoting Y amount of dollars for my writeup or as a grant to me more broadly, or someone will seriously consider working with me full-time.
I think the US intelligence community has had absolute unaccountable power to take key decisions of the US govt for the last 80 years (since world war 2), and I don’t see this changing by default.
Key decisions they unilaterally take include decisions on sanctions (on equipment, reagents, etc) to stay at the frontier on any technological race in last 80 years. Another key decision they unilaterally take includes decisions to start wars.
They are not accountable to the Congress / Senate, or to the Supreme Court, or to any regulatory body or similar, when it comes to these key decisions.
Heads of US intelligence already have active relationships with the heads of all major US AI companies. They are one of the primary reasons the US AI companies will remain unaccountable to the US public, unless something big changes.
I think social media influencers are most accountable to the public. Politicians and billionaires are less unaccountable than them. Heads of intelligence are the least accountable to the public. There are structural reasons why this unaccountability exists.
Persuading heads of US intelligence community to stop the race towards ASI is too hard, they are likely to accelerate towards ASI even if they have some understanding of the risks.
I have a fairly strong belief in this direction. Why?
I have studied every US govt whistleblower leaking classified info in the past 50 years as part of the work I published.
I have read atleast a couple books about who actually took the decisions during the Manhattan project.
I have read a bit about US cold war history, and a bit on more recent wars started by the US, to see who took the decisions.
I have also looked at some of the journalists, politicians and mass protest movements in the US trying to (unsuccessfully) hold the US intelligence community more accountable, in more recent times.
I have read the blogs of almost every famous cypherpunk activist, and have carefully tracked who got arrested versus not, and why.
I have gone through public statements of many recent heads of US intelligence on topics including AI, cybersecurity, social media, surveillance, drone warfare, and so on.
(Incomplete list) Examples where I think the US intelligence community was held even a little bit accountable to the public
I think the Vietnam war stopped sooner than it would have otherwise, because of mass protests.
My guess is that one factor for why Truman didn’t support von Nuemann’s plan of doing a pre-emptive first strike and creating a nuclear monopoly for the US, was that he was afraid of mass protests and the authoritarianism he would have to engage in to suppress these protests.
How can you change my mind here?
The best way to change my mind is obviously to actually get US IC members in support of an AI pause.
I have very little hope in the “inside game” strategy that EA billionaires are attempting, such as putting Jason Matheny in RAND or Paul Christiano in NIST or similar.
I have slightly more hope in being Loud and Annoying. Actually go organise protests at Fort Meade. Actually go figure out which social media all these US IC people use and persistently call them out there, for not taking a stance on AI pause.
The second best way to change my mind is to offer historical evidence like I have shared above. Although, given that I already have a large number of datapoints on my side, just one datapoint on the opposite side won’t be enough, you’ll need either many datapoints or a datapoint with a lot of weight.
This section could be an interesting article. I mean, listing the specific examples.
Ugh. Fair enough
Short answer
US govt whistleblowers—Read my US govt whistleblower database
Manhattan project—Read the wikipedia pages, read Making of atomic bomb by Richard Rhodes. Just go ask some historians in academia, I think what I am stating here is the consensus viewpoint, and is not that controversial or politicised.
US Cold War history—This stuff is a bit more politicised by the leftists (partly because counterculture movement took off in 60s). You can read Devil’s chessboard by David Talbot for left-leaning view on this.
Journalists, politicians, mass protest movements against the US IC today—I don’t have any good resources on this, but you’ll easily get plenty of (extremely politicised) sources on No Kings and anti-ICE protests and so on. Both left-leaning and right-leaning groups in US have organised various protests, there’s a long list of YouTubers and so on. Ask gpt-5.5 to compile it.
Cypherpunk activists—I probably do have the list of cypherpunk activists compiled somewhere in my notes, but I can’t find it at the moment. Again just go ask gpt-5.5, and actually go read the original blogs, not what the media wrote about them. Or go read the cypherpunks mailing list itself. This index has organised it by popular authors. The short version is that people working on encryption (PGP, Monero, Zcash, Tornado cash, NIST crypto standards) and leaking of classified information (Wikileaks, Blacknet, etc) are the ones who have been the most relentlessly targeted and imprisoned. If you’re just working on a secure OS or on blockchain more broadly or on dark web cloud infra more broadly, you’re less likely to be targeted. There is a long history here. For instance, civilian usage of encrypted radio was banned during or after world war 2, after both Axis and Allies used Numbers Stations to send messages, and it remains banned to this day. Digital encryption was initial regulated as a munition, but due to courage and effort of the PGP authors and a few other cypherpunk, PGP atleast exists now and is legal to use. Almost nobody uses it, and people working on such tech are at higher risk even today.
Public statements by heads of US IC in the last few years—Read my Theory of change for US govt whistleblower database and guide, in particular the red-teaming section
Why I am not writing a long answer
I am busy doing actual object-level work, and I have already wasted a year of my life arguing on lesswrong with random people with no skin in the game. One year is extremely costly given how close we are to the end of the world.
If someone who actually has expertise or power can commit some actual skin-in-the-game, I can spend more time talking to them. For example, if someone says they will pre-commit to devoting X hours of their life giving feedback, or someone will pre-commit to devoting Y amount of dollars for my writeup or as a grant to me more broadly, or someone will seriously consider working with me full-time.