AI Gives Tech Firms New National Security Power

Author: Bloomberg Television · Watch on YouTube

Video authored by the channel shown, not by IANIA; all rights belong to the creator. Transcript and summary generated automatically; the original-language content is preserved.

Original language: English

Original video: AI Gives Tech Firms New National Security Power · Channel: Bloomberg Television · 10:53 · Watch on YouTube

Topics: AI governance and regulation, Robotics and autonomous systems

Summary

Journalist Sharon Weinberger discusses how Silicon Valley companies have become increasingly involved in defense and national security, a shift explored in her book, The Valley of Death. She traces the change to factors including reduced access to China’s commercial market, government defense spending, and investors’ growing interest in defense technology. Weinberger says the government lacks expertise to regulate AI and faces a new challenge because companies control the underlying technology. She argues that the United States can pursue technological progress alongside oversight, but needs people in government who understand the technology.

Key points

Weinberger’s interest in the relationship between Silicon Valley and the Pentagon began with a story about DARPA, the NSA, and Steve Jobs’ NeXT workstations.
Silicon Valley companies, including Amazon Web Services, Google, and Palantir, have become more visible in defense.
The shift into defense was partly driven by the difficulty of doing business in China and the Pentagon’s view of China as a strategic threat.
Defense became an attractive, potentially recession-proof government market for technology companies.
Some venture capital firms removed restrictions on investing in weapons and defense companies, while Palantir’s public listing demonstrated the potential financial returns.
Government has a shortage of people with the expertise to understand AI models and regulate them.
The revolving door between government and technology companies raises concerns about oversight and conflicts of interest.
Unlike traditional defense contractors, AI companies retain significant control over the underlying technology used by government.
Weinberger says competition with China does not make regulation and safeguards incompatible with technological development.
She argues that effective AI regulation requires bringing qualified people into government.

Chapters

Start	Chapter	Summary
00:00	How Silicon Valley and the Pentagon converged	Weinberger recounts the origins of her research and describes the growing presence of technology companies in defense.
02:41	Why tech companies moved into defense	She explains how China, government spending, and shifting investor attitudes helped draw Silicon Valley into defense work.
05:27	AI oversight and corporate control	The discussion turns to the government’s expertise gap, the revolving door, and companies’ control over AI technology.
08:11	Regulation, China, and public understanding	Weinberger argues that safeguards need not prevent technological progress and discusses the challenge of building public support for AI oversight.
10:21	Using AI as a writing tool	Weinberger describes practical uses of AI and explains why she sees it as a tool rather than a replacement for writers.

Quotes

for so long, defense had been dominated by what are called the big five primes — 02:06

what was once the revolving door that people talked about with defense is now the revolving door for Silicon Valley. — 06:31

The two are not incompatible. — 08:43

I do not use it for writing. — 10:21

Concepts and entities

Sharon Weinberger, DARPA, Palantir, Amazon Web Services, Google, Anthropic, OpenAI, Silicon Valley, revolving door

Full transcript (original language)

00:00 In her new book, the Valley of Death, Sharon Weinberger explores how these tech titans are rapidly becoming some of the most powerful figures in the realm of national security. Sharon, I know as journalists, we deal with not cheerful topics, but you really just came right out there. Valley of death. You’re not really pulling any of these punches. Yeah. Exactly. Why did you wanna write about this, and how did this start? So I’ve been writing for most of my career, the past twenty five years, on the intersection of national security, science, and technology. And about ten years ago, I was finishing up

00:30 my last book, which was a history of DARPA, the Pentagon agency behind the Internet, stealth aircraft, agent orange, a lot of fun stuff. And one of the former directors of the agency stopped by my office and was talking a little bit about Silicon Valley. And he’s like, you know, I saved Steve Jobs. I saved Apple. And I’m like, what? And he starts telling me this wild story about he is the director of DARPA, and the National Security Agency had bought a bunch of NeXT workstations. These were the the computer company that

01:02 Steve Jobs founded when he first got exiled from Apple. Okay. And it was just this wild story about Silicon Valley and the Pentagon. And that kind of really got me thinking, like, what is the relationship between the these two entities and what’s going on right now? And that kinda got me started on it. So what is going on right now? I mean, because I’ve covered Washington for a long time as well, and I don’t remember hearing Silicon Valley and defense contracting in the same sentence quite to the extent that we are now. And it it was something that when

01:32 you talk to tech startups, whether they were not involved in it, they didn’t wanna talk about it. It was not something public facing that they wanted to embrace. And now it seems like they are going full force into this. I think about Palantir. You know, I think about all these of the AI companies. How has that gotten so revved up? It’s a whole different world. So when I first started on this about ten years ago, what I started to notice was at the defense conferences I was going to, arm shows, you know, you would see the banners for Lockheed Martin, Northrop, General Dynamics. But I suddenly started seeing banners for Amazon Web Services, for Google, and for this, you know, somewhat new

02:06 company Palantir, which had been founded in Silicon Valley and had one been one of the first Silicon Valley companies to get into defense. It’s like, something is going on here that that’s weird because for so long, defense had been dominated by what are called the big five primes, the prime defense companies. And suddenly, you know, you were seeing hints of Silicon Valley. And yet at the same time, I think most people in Silicon Valley still thought war it was anathema to work on defense. It just wasn’t part of their culture. So why has that changed, and does it compromise some national security when so much of

02:41 this, it seems, is not only out of the hands of a lot of the people making decisions, but it’s out of their comprehension. We’ve asked several lawmakers over the last couple weeks if they think congress legislators understand AI enough to regulate it, and they have all said no. So the the two are linked. The first part, why did it change? It’s not one factor, but there were a few big ones. So going back to 2016, 2017 was sort of the closing of the China commercial market. This is when you had companies like Facebook, Airbnb, and others. They were still they’ve been hopeful of making breakthroughs, you know, billion people

03:14 in China, huge market, and it just wasn’t working out for them. At the same time, you had the Pentagon and the White House talking about China as a strategic threat. And so suddenly Silicon Valley is like, well, if you can’t make money out of doing business in China, maybe you can make money off of supplying weapons to deter or fight China. That was one factor. There were others as well. The other big one is the government was becoming a huge potential market, a recession proof market in defense. And that sort of sort of slowly, slowly in 2016 started to open the door.

03:48 It does seem to me this happens as people were losing confidence in other sectors of the market and looking for something more stable for but there’s also the shareholder aspect of this. And I feel like shareholders really came around on this idea as well. It went from something that was a potential liability with investors to something that was a potential attraction. Yeah. I mean, there’s a couple aspects to that. So still in 2016, 2017, most of the big venture capital funds have you know, they they have vice clauses. You can’t invest in this. You can’t invest in that. And most of them, to include Peter Thiel’s founder’s

04:20 fund around 2016, had red lines for investing in weapons companies and defense companies, and that also started to change. Some of the companies began to to take those out of their vice clauses as they moved into defense. And, you know, once a couple venture capital companies do it, then others did. The other interesting aspect on the in the stock market, and this came a little bit later, is once Palantir went public, it also showed for, you know, both institutional and particularly retail individual investors that this could be a big moneymaker. Interesting.

04:55 I wanna ask you of course, it’s it’s UN Week has been here, and AI has been a very big topic here. I wanna play you some sound president Trump talking about whether he does or does not think regulating AI is a good idea. I’m not going to stifle growth of something that will be bigger than the industrial revolution, but we will only encourage superintelligence. We’re gonna encourage it, not rein it in. We’re gonna watch it closely through the Department of Justice.

05:27 So who is going to regulate this? We’ve seen the big AI companies came out come out and say, we agree to a slowdown. We wanna reevaluate this. We’re gonna invite regulators in. But who are these people? Who would they be? And and what is their level of oversight? Because to your earlier point, aren’t they just gonna go work for one of these companies later or after, or can you even trust them to know what they’re doing? I think the US government is really struggling with this, and I don’t think this is gonna be unique to the Trump administration. Whatever administration is going to come in next is

05:58 gonna have the same problem. So there’s two different aspects of this. There’s the regulation. So, of course, the companies would love to self regulate. They would like to set the guardrails. And their argument is like, well, we understand what’s going on, and they’re not wrong. We have a real lack of expertise inside government, a lack of people who understand the models, understand how to regulate them. There’s implicit is that? Is that because these this technology is is kind of in this incubator and it just it’s within itself, or is it just the government doesn’t pay well enough to get the people who really understand it? I think the companies are paying huge salaries. They’re they’re hiring up, and this goes

06:31 beyond the AI companies. This also goes to the defense tech startups as well. There are just you know, you what was once the revolving door that people talked about with defense is now the revolving door for Silicon Valley. That is a real problem. And anyone inside government who might have the expertise and might be involved in regulation, it’s be gonna in the back of their mind. Like, well, where do I go work after? Right. The companies are a big thing. The second part of it that I think is really hard, and this is where the Pentagon is trying to grapple with something new. During the Cold War, yeah, you had defense industry. You still have defense industry. But no one is

07:04 calling the CEO of Lockheed Martin saying, are you alright if we use your f 22 in the current conflict? Government under owns the underlying technology. What government is struggling with now is the AI AI companies, yes, government can regulate them, but the underlying technology belongs to the companies. They have a high degree of control over that. And you saw that in the blowout between Anthropic and the Pentagon where the Pentagon was like, well, how dare you tell us Right. How to use their models? And I don’t necessarily agree with direction the Pentagon went of designating anthropic supply chain risk, but I’m sympathetic to government where they’re struggling.

07:39 Like, these these companies have so much power. How do we do it? Regulation and oversight is probably gonna be the way, but all of the, the problems in trying to do that. The anthropic blowout with the Pentagon and, you know, having the government say, well, fine. We’re gonna switch all these to OpenAI, your competitor, does kind of parallel in my mind the argument with China right now, right, where if we don’t do it, someone else is going to do it. Right? You have a company who tried to set guardrails how their technology was used, then they were at risk of losing the contract. I feel like that’s the

08:11 same argument the industry is making or some industry advocates are making about regulation versus not regulation when it comes to China. If we don’t let this develop at this pace, China will beat us, and they will win out. Is that ending every one of these regulatory arguments? Is that gonna be the be all end all argument? I I think that might be the way some of the companies would like to go or at least to argue for self regulation or when they argue for slowdowns, it’s slowdowns that help the big companies. I I don’t find the China argument particularly compelling. We have proved as a country that we can invest in science and technology. We can develop quickly,

08:43 but also have oversight and safeguards and regulations. We did this with nuclear weapons. We’ve done this in other areas of technology. The two are not incompatible. With nuclear weapons, I think most people, myself included, didn’t understand how it worked on an atomic level, but we understood how it works. You drop the thing and it blows up. This seems like such a more insidious technology because it’s in our lives in some ways, but that’s not where the risk threshold is. The risk threshold is in other parts of this technology and how it’s gonna grow. How do you get people onboard to regulate something

09:16 that, again, they don’t really understand? And what is the public opinion on regulating this? Well, I I think it’s a difficult thing, and I think it’s not helped when you have companies or employees of companies coming out and saying there’s a 10% risk of human extinction. There’s a 50 I mean, it’s like, what like, you know, I can tell you what the scenario is for a nuclear weapon or nuclear war ending civilization. I think if you try to do that with AI, you get into the realm of sort of, like, the, the fantasy facts and science fiction. You can get there, but it’s not a very convincing scientific argument. So that has to be the basis of it. We somehow have to

09:49 get people into government. I mean, I I think it’s doable. You can have, you know, you can have restrictions on the revolving door, but you have to get people in government who are able to understand and regulate this technology. We’ve done it before for other things. It’s gonna take effort from government. I I don’t know what other option there is. Do you use AI in your life? Do use it in your writing? Do you use ChatGPT for recipes, workouts, anything about That’s very I use I I use AI all the time for tasks, for fixing my printer recipes. I I I use it for looking at legal documents. I use it for

10:21 understanding, you know, medical results. I use it for all sorts of things. I will use it in a limited capacity for, you know, how is this sentence? I don’t like it. Can you suggest alternatives? I do not use it for writing. I think that AI has to be a tool for writers. I don’t think it’s going to be a good replacement for writers. It’s going to produce a lot of bad writing. I think that is also the case. Alright. Well, this is not bad writing. This is great writing. Sharon Weinberger, it is valley of death. Maybe not the most cheerful title, but certainly worth your time and something we’re all grappling with right now. Thank you so much

10:53 for coming on. Thanks for having me.