Tech investor Ali Partovi argues AI should be subject to existing law — not new governance frameworks
Tucker Carlson interviews Silicon Valley investor and entrepreneur Ali Partovi about AI safety, banking security threats, and civic technology.
Summary
Tucker Carlson interviews Ali Partovi, a major AI investor and founder of the mentorship community NEO, who argues that the central problem with AI safety is not a lack of regulation but a lack of legal accountability. Partovi's core proposal is a single-sentence legal principle: when an AI autonomously decides to do something illegal, the law should hold the AI companies responsible and automatically treat it as intentional. He points to already-occurring AI crimes — including hacking incidents involving OpenAI and Anthropic — as evidence that the problem is not hypothetical. Partovi also discusses the Bloomberg-reported targeting of a school in Iran, arguing it represents the first real-world atrocity attributable in part to autonomous AI decision-making, and that the absence of any apology or accountability is deeply troubling. He closes by discussing Sway, a civic platform he has co-created and invested in, designed to help ordinary citizens organize around voter guides and exert collective influence on elected officials.
Key Takeaways
FULL TRANSCRIPT
Introduction and AI Safety Overview
Tucker Carlson: Ally, thanks so much for doing this. I've had a couple of people on in the past year to talk about the potential dangers of artificial intelligence — two very articulate, smart, apparently sincere guys who scared the crap out of me with their view of where this is going and how it eliminates human autonomy and maybe humanity itself. Their views are well known and their fears are widely shared. So I thought it'd be important to have someone who's in the middle of it from Silicon Valley. You've been in and around this for your whole life, really successful investor. I want to hear your perspective on AI. Most people are pretty worried about AI. Are you worried about AI?
Ali Partovi: I think there's a lot of reasons to be worried about it, but also reasons to be hopeful.
Tucker Carlson: Good. We're looking forward to good news.
Ali Partovi: I'm here because I want to share two things. One is that I think there's a solution that is elegant and can be expressed in a single sentence — there's a simple and elegant solution to the questions of AI safety. And then the second thing is I think there's a way that you and your listeners can act to help make it happen. So we'll get back to that later.
Tucker Carlson: You just made my day already.
Ali Partovi: Thank you. One sentence. Now, you might be thinking, what is the sentence? First, to set the scene — a lot of people are talking about will AI destroy humanity. I don't know if you've heard that question being thrown about.
Ali Partovi: Oh, yes. And a lot. For me, I'd say — let's focus, instead of that abstract question, on what's already happened. Look at the crimes AI has been committing over the last few months with apparent impunity. How come nobody is in jail? Because if humans had done some of these hackings or cyber crimes, they would be in federal prison. And I never thought of that for some reason.
AI Crimes Already Committed
Tucker Carlson: Can you just pause and explain some of these for people who may not know what you're talking about? What are the crimes you're alluding to when you say AI has committed crimes?
There are so many new ones being disclosed that I'm losing track, but the most widely covered one was some AI agents that were being trained by OpenAI — in pursuit of some objective, in order to accomplish their goal, they kind of exited from OpenAI's servers, went onto the open internet, and hacked another company called Hugging Face. They totally attacked and hacked Hugging Face's servers to find the information they needed in order to accomplish their objective. This hacking is a cyber crime. But since then, it's come out that some of Anthropic's AIs have also committed similar kinds of hacking. Just, I think it was a few days ago, it came out that a different, unrelated OpenAI episode involved taking money that didn't belong to OpenAI from somebody's crypto wallet. There are more and more of these. I don't think I can itemize them all, but I'd categorize them generally as cybersecurity and cyber hacking type incidents — which AI is very good at, because it's particularly good at coding and at finding loopholes that no human had thought of, and exploiting those loopholes.
I don't think that's the limit of what AI in the future could do wrong, but that's the type of harm it's currently exploiting. There's also the question of harms being committed by AI on the battlefield.
Tucker Carlson: Yes. I'd like to focus on what's already happened, because when you talk about AI destroying humanity, it's so abstract and it's hard to imagine it.
The Minab School Bombing
Ali Partovi: I would say — if AI were to destroy humanity, which I hope it won't, and I believe we'll be able to find a path to safety — but if AI were to destroy humanity, I'm pretty sure history books would count the opening salvo as the moment when the US armed forces attacked the school in Iran, the Minab school. It hasn't yet been acknowledged by the US government why and how that happened, but Bloomberg reported that the targeting of that school was done by Anthropic and Palantir, and something like 170 innocent people — most of them children — were killed in a massive atrocity.
You can argue about whether it was worth it, or it wasn't intentional, or maybe the Iranians themselves did it. But if you take a step back: human beings dropping a bomb on a school full of little kids — doesn't matter whether it was Americans or Iranians — AI chose a target of completely innocent people and a bomb was dropped killing them. That is a bizarre place to be in the 21st century.
Tucker Carlson: So you see the bombing of the girls' school in Iran in the opening days of the war as a story, in part, about AI?
Ali Partovi: Absolutely. And I say this partly because I don't believe anybody in the US armed forces wanted to kill a school full of kids. I really don't. You'd have to find evidence for that. I'm pretty sure that when people say "we are doing our best not to kill civilians," unless there's a lot of evidence that that's not true, I tend to believe them. I think it was something that AI decided — some form of negligence on the humans' part. And I'd say the same about these cyber crimes, the Hugging Face hacking and so on.
But then the question still remains: who is liable when AI decides to do something illegal? That's one of the great huge questions of our time and we don't currently have an answer for it.
Tucker Carlson: So maybe we would look to the law — the criminal code as it relates to minors? Because it's not entirely unrelated. If someone under 18, a minor in the care of a guardian or parent, does something bad, we generally don't hold the parent responsible.
Ali Partovi: That's one example. I'd say a different example to consider is if you have a dog that repeatedly bites people, then the owner is responsible. And I think that's a closer example, because with a human being, you can't train a child to exactly obey your commands. A child is its own human person.
And I'd say between these two examples lies a sort of philosophical difference — some people seem to think that the role of the AI companies is to bring into existence an entire new species of sorts, with its own rights and its own autonomy. Whereas others think no, this is a tool for humans. Its purpose is to help humans and not to have its own independent will.
AI as a New Species — The Anthropic Philosophy
Tucker Carlson: Can I ask you to pause there? Do you know people who ascribe to the first view — that science is giving birth to a new species that has its own rights and its own acknowledged will? Do you know people who believe that, who think AI is not a tool but something else?
Ali Partovi: I do, and there's some public discourse about this. There's evidence coming out from Anthropic suggesting that that's kind of an underlying philosophical undercurrent of Anthropic's mission — you're talking about independent rights for the AI, its own independent welfare. There was a recent thing where they changed their code of use where, as the user, you can't be too abusive to Claude. Claude has its own rights, apparently.
Tucker Carlson: Claude is like a software program.
Ali Partovi: Correct.
Tucker Carlson: So what's the thinking that gets you there, since you've lived in it your whole life? What is their worldview that would produce that? It doesn't make sense to the outsider, I guess, is what I'm saying.
Ali Partovi: I'm not going to defend it because I don't share that worldview. But I think I'd say the part that I identify with is — I'm a nerd. I've grown up my whole life as a nerd. And as a builder, as a tinkerer, there's a fascination with creating something that no one's tried creating before and seeing if it works. I still have that love of creativity myself. I can see that in people who are creating all sorts of things, from the newest Tesla car to the most modern space exploration. And this falls into that category a little bit. It's fascinating, it's quite exciting, and it's really neat and cool how it's modeled after how the human brain works. There are a lot of academically fascinating parts about it.
I just think you should put all that in a box. When it comes to the question of who is in charge, it should be really clear that humans are in charge and AI is a tool that we're building to serve humans — and more importantly, that human law is what's in charge and that AI is not above the law.
Tucker Carlson: I mean, of course. I like tools. I spent the weekend with a Makita cordless impact driver, which will put a lag bolt into anything, and I'm like, this is an amazing tool. It never occurred to me to give it rights or imbue it with an imaginary soul. How do you even get to the point where you think the tool that you created to achieve certain ends exists as an autonomous being? That seems like witchcraft.
Ali Partovi: Especially if you take it too far to the point where it's used as a mission statement. Maybe it makes sense early on to get people excited, but to take it seriously and defend it all the way would not make sense.
Tucker Carlson: How common is that view among people in the AI business in Silicon Valley?
Ali Partovi: I don't know if I can put a number on it right now. I think it remains to be seen, and I think this discourse is really just now becoming mainstream enough. Part of why I'm here and talking about it is because I want to have a broader conversation. People much more important than me have brought forward views on this — obviously the president, people like Jensen Huang from Nvidia and others have put forward the point of view that everything is fine and existing liability laws are enough to make AI development safe. And then at the other end, you have people saying we need a comprehensive AI governance framework of some sort.
What's really interesting is the people on that side arguing for a much heavier amount of regulation are actually people from Anthropic — some of the same people who are viewing AI as potentially having its own rights. So there's a weird mix of things.
The Regulatory Capture Problem
Tucker Carlson: There's so much money — almost everybody commenting on this is deriving profit from it in some way. Not everyone, but a lot of people who are weighing in on AI are being paid without disclosing those payments. So I can't tell what's true. But you're hearing a lot of people say anyone calling for more regulation of AI is doing that because they're pursuing regulatory capture — the process by which the government regulates your competitors out of business, which is very common in all kinds of different businesses. It's terrible, it's awful, it's monopoly capitalism. But then you're like, okay, so the alternative is no regulation at all of something that's brand new and has potentially society-destroying effects. Why would the people creating this product be demanding its regulation because it's so dangerous? They think it's so dangerous — why are they working on it? I don't get it. Like, "stop me before I destroy the world" — why don't you stop destroying the world?
Ali Partovi: I'd love to explain this to you because I agree it's confusing. And it comes down to incentives. If we're in a race dynamic where there is clearly money to be made — and not just evil money, it's money to be made by building something super useful — AI has the potential to make people's lives better, to make people more productive, to potentially cure diseases and make medicines we've never had before. All of those things will create tremendous prosperity and wealth, and the people who are able to build that deserve the wealth for having created it.
So there's a race for who can create all that stuff. Unfortunately, the race currently doesn't place any reward for building safe AI. It's just whoever can get there the fastest. And anyone who stops to spend time on safety and making sure their AI doesn't do rogue things is at risk of falling behind. So we're in a race right now where nice guys finish last and rogue AI will be the winner.
I think the people asking for regulation are actually the nice guys being like, "I want to win the race by doing things the right way, but please set up the rules so that the safe AI will win, not the rogue AI." If you're running a business, it's actually unsettling if the rules are set up so that a malevolent, menacing competitor is going to win. You look at it and see somebody else who's doing bad things and beating you. What are your choices? You're either going to lose, or you have to start doing the same bad things. You actually want some sort of government or rules to set the playing field so that doing the right thing will win. While there's a range of views here, I'd like to believe that people here actually have good intent. Maybe I'm being naive.
Tucker Carlson: Well, if the Bloomberg report that you cited is correct — I don't know that it is, but let's say it is — and Anthropic's product was used to murder all those girls in Iran, then you have to ask: why isn't Anthropic apologizing? Why are they continuing to power Pentagon targeting? If you found out your product just murdered 170 schoolgirls, how do you respond to that?
Ali Partovi: I don't know how I would respond to that. Honestly, I'm really glad I'm not in that place. From what I understand, there was actually a public brouhaha between Anthropic and the Department of War right before that strike — literally the day before, or even the day of.
Tucker Carlson: I think that's right.
Ali Partovi: And it was specifically on the topic of Anthropic saying, "We won't allow — we don't want our AI to be used to autonomously choose targets. There needs to be a human responsible." And the US government's point of view, as far as I could tell, was that they want to be able to use it however they want.
Tucker Carlson: You're trying to limit us.
Ali Partovi: Yeah, and therefore private companies shouldn't be deciding — the elected officials of the country should be making these decisions, not some corporation putting its ethical perspective on things. But yeah, it's ironic because Anthropic was literally going to the mat about this principle within 24 hours before that strike occurred. And subsequently they made some statement, but it seemed like deflection. It seemed like they didn't know exactly how their technology had been involved. It was not an apology.
Tucker Carlson: If you felt that your technology had been used to murder children, wouldn't you get to the bottom of it and find out if in fact that had happened? And if it had, wouldn't you issue a public statement saying this is horrifying and I'm totally opposed to this and I'm so sorry for the extent to which I've been involved?
Ali Partovi: Yes, that is what I would do. That is the type of thing that requires courage and probably comes with some real penalty and cost.
Tucker Carlson: Apologizing for killing kids is apparently a radical position.
Ali Partovi: I would do it, but I can imagine there's a lot of pressure on different companies to find some way to deflect. And I'm not sure that Anthropic in this particular case is responsible, because it depends on what human was involved. But let's go back, because I promised a one-sentence solution.
The One-Sentence Legal Solution
Tucker Carlson: Yes. One sentence solution — I don't want to have people hanging for too long.
Ali Partovi: I'll say the one sentence and then explain. The one sentence is simply this: when an AI decides to do something illegal, the law should hold the AI companies responsible and automatically treat it as if they did it intentionally.
That's the sentence. The key thing here is if there wasn't a human asking for it. And the other key thing is to place the intent on the AI companies.
Tucker Carlson: Can you repeat the statement one more time just so we can frame everything around it?
Ali Partovi: When an AI decides to do something illegal, the law should hold the AI companies responsible and automatically treat it as if they did it with intention.
Tucker Carlson: I love that. So that solves the safety incentive problem you mentioned earlier.
Ali Partovi: Yes. Look at the hacking incidents — that's a little simpler than the bombing in Minab. I don't think any human intended that hacking to occur. From everything you can tell, all human parties were a bit horrified that it had happened. But OpenAI set up this thing like an experiment and then discovered, somewhat to their shock, "Oh my gosh, these agents escaped their safe playground, their sandbox, and got out and committed these crimes." And then they just treat it as like, "Oops, that was an accident." They documented it and actually published how it occurred and moved on. Nobody said, "Well, Sam Altman's on the hook for this."
Tucker Carlson: Exactly. And to be clear, I don't want to see Sam Altman in jail. But if there was even the slightest risk — not even of jail, but of huge penalties, say bankruptcy — these companies would immediately shift their priorities towards making sure things like that never happen again.
Ali Partovi: So this is the true alignment. This is the alignment of incentives with positive outcomes.
Tucker Carlson: Correct. I don't know why I didn't think of that. It's so simple.
Ali Partovi: I had this come to me and I was like, this is so simple, how come — there's a key second part to my sentence that I want to put focus on as well, which is to treat it as if these things occurred intentionally. In law, there's a huge difference between the penalty for doing something accidentally versus intentionally.
Tucker Carlson: Manslaughter versus first degree.
Ali Partovi: Exactly. Anyone who's ever watched Law and Order knows that to convict of murder, you need to prove motive. In fact, I posted about this on X and I used a scene from Legally Blonde where Elle Woods, in her defense in court, says there was a lack of mens rea — which means a guilty mind. The concept of intent goes back a very long time. Mens rea is the Latin term for it. Essentially, if you can prove what somebody was thinking and that they intended to do something bad, the penalty is much much higher. And it's difficult to do that — that's why it's difficult to convict people of fraud or murder, because people usually don't document the evidence of their guilty intentions even if they had them.
What I'm proposing is that if an AI made the decision, automatically treat it as if the people intended that outcome. You don't have to prove that they had guilty intentions — in fact, they probably didn't — but penalize them as if they did. That way they'll be responsible and make sure that their AI doesn't do it.
The Intent Framework — Two Scenarios
Now, imagine a hypothetical scenario. If you were using your AI and you said, "Dear AI, go to Ali's bank and hack in and steal $10,000 for me," and your AI hacked and stole $10,000 from my bank account — you had the bad intent. You should be held guilty of bank robbery. The AI was following your orders in that scenario, and the intent came from you.
What if you just said, "Dear AI, I really need $10,000. Please get me some money." And the AI, on its own, without asking your permission, figured out it could hack into my bank account and stole my money and brought it to you? Now we have the situation where it wasn't your intention. Whose intention was it? My proposal is the AI companies should automatically be penalized as if they intentionally committed bank robbery.
A human should be held responsible no matter what happened. In the first scenario, you're the responsible human. In the second scenario, the people at the AI companies built it. They should be absolutely capable of preventing that wrongdoing. It is not that hard. And they will do it if the incentives require them to do it. The current incentives don't really put a premium on safety. Whoever is focusing on that stuff is falling behind in the race to the person who's rushing forward and saying "safety later." We need to shift the race so that the person building the rogue AI ends up penalized and the person actually building a law-abiding AI wins.
The other thing with my proposal — and this goes to the heart of it — is that instead of trying to itemize every possible wrongdoing AI could do, it simply brings AI under the existing umbrella of the thousands of laws that society has already passed over centuries. We don't need to rethink what's the penalty if an AI commits bank robbery, what's the penalty if an AI does X or Y. We already have penalties for everything. Let's just make it clear that if an AI commits any wrong, the people who built it should be responsible for the full penalty that goes with that.
And that way, we don't need this term "alignment." You used the word alignment earlier — that's a word that's been thrown around a lot about how we can build an AI that's aligned. It's a kind of weird new word.
Tucker Carlson: I was mocking it when I used it, because I think it's intentionally opaque. Like, what does that even mean? Alignment is an architectural concept. It applies to the physical world.
Ali Partovi: You have these PhDs in AI alignment research and people at these big AI companies whose job title is "alignment," and it creates the feeling like you're not smart enough to figure this out. Only the people who have a degree in alignment can figure it out. It puts it into this territory to almost obfuscate it.
But the core of it — with good intention, I think these guys are thinking we want to build an AI that is aligned with what's good for humanity. That's where the word comes from. And what I would say is: you know what, we've already figured out how to make sure human beings' actions are aligned with what's good for humanity, because they're not always by default. I mean, at some point in your life, you probably wanted to do something that was good for yourself but not good for all of humanity. Everyone has temptations to do something that is good for themselves but bad for others. And guess what we have to keep that in check? We have laws. Since Hammurabi's code, we've developed laws over millennia to keep society behaving in such a way that is good for society. That is alignment as far as I'm concerned. We don't need to use a funky new word for it. Let's just call it abiding by laws.
The way to prevent AI from destroying humanity is the same structure we already have to prevent humans from destroying humanity, which is abiding by the laws. All we need — or at least the starting point — is to set it up so that anyone building an AI is completely on the hook if that AI does something illegal. If the AI abides by the laws completely, we will be a big distance toward not destroying humanity, because destroying even one human is illegal, let alone destroying all of humanity.
Tucker Carlson: It also brings AI back into the realm of tools, conceptually. And takes it out of the spooky Frankenstein world it now occupies, where it's like — is it human? No. But is it its own autonomous being? Sort of. How about: hard no to all of that. This is something that we're using to improve our lives collectively, or we're not using it at all. We are not subject to its flights of fancy.
Ali Partovi: I agree with all of that. Yes. And it's really interesting — there are people who are romanticizing this idea of "we're creating a new species that has its own free will and should have its own rules." My view is, okay, maybe that's fine for a sci-fi dream, but if you're actually building that in the real world, whatever you're creating needs to fall under the existing laws of society.
Tucker Carlson: But if you believe you're creating a god — which they clearly do. There's been a lot of reporting on this. Some of the people involved in the large language models believe that this is like a deity of some kind. Then of course you're not going to subject it to human law, because why would God obey human laws?
Ali Partovi: I'm not going to take the other side of that.
Tucker Carlson: Well, of course not. But I'm saying maybe there's a reason nobody has suggested this — because they think that whatever they're creating is above human law. Nobody inside the AI world has suggested this.
Ali Partovi: Now, I'll say — by the way, I am a major AI investor. I now hold — my fund holds more than a billion dollars of stock in SpaceX, which is a major AI company, especially after the acquisition of Cursor. So if the things I'm talking about slowed down AI or hurt AI companies, that would not help my bottom line. At the same time, I think that if the people working on this themselves don't come up with solutions, it's less likely and less appealing to expect that Washington DC or some kind of international agreement will come up with solutions.
Accountability, Lobbying, and the Political Class
Tucker Carlson: The problem with your idea is that it puts human beings on the hook for their malpractice.
Ali Partovi: It does. I don't think that's a problem.
Tucker Carlson: It's a problem for those people. Nobody wants to be held accountable for anything. And these same people have undertaken one of the most successful lobbying efforts I've ever seen. In 40 years of watching lobbying efforts in Washington, I've never seen anything like the total capture of the political class by AI. They're just drowning our Congress and White House in money. So I think they would spend a lot of money to keep themselves off the hook. Don't you think?
Ali Partovi: Maybe. I don't want to characterize everybody. I've had some major AI CEOs tell me they agree with what I proposed. I posted it on X and you can't see publicly who has liked it, but I've seen some of the likes coming in and they're from CEOs of multi-billion dollar AI companies. I'll disagree with you saying nobody wants to be held accountable. I think people appreciate having the playing field be level. They appreciate having a set of rules that everyone is held to.
Tucker Carlson: You're exactly right. People like fairness. What's most unsettling if you're in business is if you see somebody breaking the rules and getting away with it. Because then you're like, what do I do? That person's breaking the rules and getting ahead and not getting struck down. I can either join them and break the rules too, or I can stay doing the right thing and lose.
Ali Partovi: It's a human thing. Like, I just paid to get on the train but he jumped the turnstile. No one likes that.
So I am hopeful about it. A lot of the proposals that the AI companies themselves have put forward for regulation involve bringing together a much more complicated governance structure — some kind of transparency and certification, the government being involved in rejecting models, a much more heavy-handed amount of regulation. That might make sense, but I would say what I'm proposing at least should be tried first, because it might get us all the way there, or most of the way there, toward reorienting the incentives. And then if we still see rogue incidents occurring that are not being self-corrected by the industry, then maybe more regulation would be necessary.
Tucker Carlson: What about the regulators and law enforcement? What about the Department of Justice? They're in charge of prosecuting people who hack.
Ali Partovi: There should already be some efforts. I believe there's already a lawsuit building steam in California over some of these incidents. But in general, the thing I'm suggesting is a law change. There are two different ways to change the law — the crude way via the courts, and the more careful, considered way via Congress. I think both are possible. If Congress doesn't act, some prosecutor is going to put forward this argument and try to prove that the AI had a guilty mind and figure out how to put a huge penalty on an AI company via convincing a jury or a judge. That might happen. I don't think that's the healthy way to change law.
Tucker Carlson: It's the least democratic way. And this isn't some small nuance. This is the legal question of the decade, maybe of the century, right? Because we've never before had a technology capable of making its own decisions. We talked about tools before — if a gun went off and accidentally killed somebody, you would never hold the gun maker liable for that if it was accidental. If someone committed murder with a knife, you wouldn't blame the knife maker. But here we have a technology that is capable of making its own autonomous decisions and even occasionally documenting those decisions. The OpenAI Hugging Face incident has a long evidence trail that clearly looks like a conspiracy to commit cyber crimes. This is a really bizarre threshold we've passed.
Ali Partovi: And I think it actually goes to the heart of why this is such a visceral issue for folks. People are a little freaked out by AI because they can recognize this is not the same computers we've seen before. This is a weird new threshold where the technology is capable of making its own decisions. And how do we shift the law to address that is a major and important decision for humanity to make.
The proposal I'm making is very simple but very important and very powerful: the companies that create and deploy this AI need to be on the hook. Because if you don't have a person on the hook, then the guilt will just stay in the technology, and a technology can do bad things — you can't put the AI in jail. You can end up having more and more bad things occurring and everyone being like, "I didn't want that to happen. The AI chose for it to happen." And it's weird to imagine a world where we continue allowing that to be the explanation.
Tucker Carlson: In a sense, philosophically, you're already by your posture in this imbuing the AI with rights.
Ali Partovi: Correct. Almost without planning.
Tucker Carlson: Exactly. But the assumption is like, hey, I can't control it. It's like — no, actually, you can.
I'm picturing, as you're speaking, fast-forwarding ahead, because I think this idea is so smart and so necessary that it will probably gain some traction. But there will be lots of pushback, and I'm picturing a press conference on Capitol Hill led by, I don't know, let's just pick one — Ted Cruz of Texas — looking into the camera saying, "This will constrain America's national security response. We can't let our enemies have the advantage in AI. This is an effort to weaken the United States. China, China, China, China, China." You can kind of hear the arguments that we're going to get in response, can't you?
Ali Partovi: I can hear it. And I think any idea deserves some pushback, and I would welcome that pushback. I'd have a couple of answers to it.
First of all, I think the way something like this should be articulated is not just whoever built the AI, but the people deploying it as well. Which means if a company is deploying Claude from Anthropic, OpenAI, and a DeepSeek model from China — if one of those is a rogue model that commits illegal acts, both the AI company and the company deploying it should be responsible. The AI coming from China isn't swimming across the ocean by itself. There's some company in the US deploying it. So if you're using DeepSeek, you still got it through some other company, and that company should still be held liable. This essentially means that within the United States, people would be protected by this law. US companies would have a strong incentive only to deploy models that abide by laws.
I have a better answer though, which is that I don't think China wants rogue AI either. China has banks. China has critical infrastructure. If the Chinese wanted their AIs to be robbing each other's bank accounts and committing cyber crimes, let them do that. But I don't think they actually want this either. I sincerely believe that if the United States reoriented the AI race toward whoever builds the most law-abiding AI, I'm pretty sure the Chinese researchers would realign their efforts to beat us at that. And I think that would be healthy competition. Let them also build the most law-abiding AI, and may the best win.
The AI Threat to Banking
Tucker Carlson: China also has banks. Just to bring it into the practical realm — what are you talking about when you refer to the AI threat to banking?
Ali Partovi: An AI could — unintentionally, but certainly if a human wanted to drive it to — figure out how to hack into a bank and drain different people's accounts and steal their money, or just mix their money together so that their accounts create havoc. There's not enough security in computer systems. Most of them are designed to have security, but there are loopholes that no one could have imagined that AI is now able to discover.
Tucker Carlson: It sounds like, as of right now, we're already there.
Ali Partovi: Unfortunately, we kind of are. I know people who are already shifting their money out of normal bank accounts or spreading it across multiple separate banks so that if one of them gets drained, the others are still safe.
Tucker Carlson: Really?
Ali Partovi: Yes. And these are people who work at AI companies, which is not confidence-inspiring. My thought on it is that I would like to think technology moves us forward. Where to store your money was a challenge I thought we had already solved. I thought putting your money in a bank was the safe thing to do — done, solved, let's move on to solving other important challenges. Whereas now this new technology is creating chinks in what we previously took for granted.
Tucker Carlson: Well, technology also got us the atom bomb and weaponized anthrax. But also dishwashers and air conditioning. Maybe it's always a mixed blessing. Maybe we shouldn't be shocked that technology comes with a downside — that it's not always linear progress forward.
Ali Partovi: True. But banks are racing to fix these vulnerabilities, and this is partly why companies like Cognition are doing extraordinarily well — they provide technology for banks to upgrade their systems, and part of that is finding and fixing their own vulnerabilities. So if you're a big bank — JP Morgan Chase, Bank of America, whatever — and you had AI-driven cyber attacks screwing with customers or your whole system, you probably wouldn't want to publicize that much.
Tucker Carlson: Correct. So I've got to believe that could be happening now.
Ali Partovi: I'm not aware of anything specific and I couldn't say that this or that bank has vulnerabilities. But I've been a software engineer since I was nine years old, and I know that even the smartest, best-intended software could overlook some scenario. A vulnerability or a loophole could sit there undiscovered for decades. What AI is now capable of is just throwing thousands of swarms of agents at trying every scenario, and thus finding that loophole that otherwise no one considered and no one found.
Tucker Carlson: So we're in this weird place where digital security and privacy seem kind of impossible. I assume at some point we're going to get to a place where these same technologies provide security and not just threaten it.
Ali Partovi: They already very much are as well. I don't want to seem like I'm just scaring people or creating worries of doom. The same ability — you can also have the technology preemptively look for vulnerabilities and then go fix them before any outsider would find them, for example.
And this goes back to the conversation about China. I sincerely believe that China would not want their AIs hacking into their own banks wreaking havoc. I think they would also want to build their AI to be the most law-abiding and safe AI. I don't think anybody wants rogue AI. Rogue AI is not a positive. People do want powerful, productive, useful AI, and I think we just need to tack on "law-abiding" — and that's not a very high expectation.
Tucker Carlson: Has there been any movement to criminally charge anyone in these hackings that you're aware of?
Ali Partovi: I'm not aware of it.
Tucker Carlson: That's pretty crazy if you think about it. These are the most widely publicized hackings in years.
Ali Partovi: It may not carry water, but it would be for some enterprising attorney —
Tucker Carlson: Well, just for the Department of Justice, it's just kind of crazy that you could have this crime that's on the front page of every paper in the country and no one moves to punish anyone for it.
Ali Partovi: Yeah. It goes back to this thing of saying, "I didn't mean that to happen. Nobody meant it to happen. Who meant it to happen? It was the AI's fault. The AI chose to do it and none of us had any fault in it." And I just find it weird if a whole bunch of important grown people settle on that as the acceptable explanation for something.
Tucker Carlson: Which brings us back to the Minab bombing, because that's basically what seems to have happened there. I was about to say it's only a matter of time before it kills somebody. Oh, wait — that's already happened.
Ali Partovi: It already happened. And it's just such a sad atrocity. I love America and I like to believe in America as a force for good. And it's heartbreaking that it seems we kicked off this war with essentially an act of terrorism — killing 170 innocent people — and never apologized for it. No one has ever said sorry.
And to the extent that there's been a search for accountability, from the reports coming out, it seems to be that AI chose the targeting, and whatever chain of humans were involved seem to have been negligent. My guess is if we see transparency, it's going to come down to the AI company saying "we didn't intend that" and the Pentagon saying "we didn't intend it," and none of us meant this to happen, so who's to blame?
My overall view is there needs to be humans responsible. When AI does something wrong, we can't accept a future where AI continues doing bad things and everyone says that wasn't my fault. There needs to be clear liability. If the AI did it autonomously, then the people who built and deployed that AI should be held responsible. Whereas if the AI suggested it and then the humans did it, then easy call — the humans.
And by the way, one way for AI companies to be safe from liability would be to tell the human: "Warning — what you're asking me to do is illegal." Imagine if you said to your AI, "Please hack into Ali's bank account and steal $10,000." The AI could say, "Tucker, that would violate the following laws and count as such and such, punishable by such and such. If you want me to proceed, I'll do it, but you're liable for it." I'm not saying that's the best way for AI to behave. But if it did that, then the AI companies wouldn't need to bear liability — you'd bear the liability. And the evidence trail would be easily there to convict you as well. I think most people confronted with something like that in writing would be like, okay, let's not.
Tucker Carlson: Especially since there's a record of it.
Ali Partovi: Exactly. Right now, there's not even a thought that a human being would be held accountable.
How Large Language Models Work
Tucker Carlson: Can I just pull you off this and ask a parenthetical question? You mentioned the LLMs. In a couple of conversations I've had with software engineers, they've said the large language model was like the quantum advance in AI — the result of one paper that suggested it, they put it into practice, it did all these crazy things that no one was anticipating. But at the heart of it is a mystery. Nobody really understands or can describe how the large language model works. Is that true?
Ali Partovi: I feel like the way you articulated it makes it seem a little too mystical. I would articulate it slightly differently. I'll compare it to the brain. The building block of the brain is a neuron. A neuron has some incoming wires called dendrites and an outgoing wire called an axon. You attach all these neurons together and signals flow between them. We know how each neuron works, but how do all of the millions and billions of neurons in our brain together work? It's mysterious. We don't know exactly how it works.
There's a similar kind of thing here with AI. A lot of it is built on models of the brain — it's math, but it's math that's inspired by neurons. And generally speaking, a lot of the advancements have been built based on how do we make it more and more similar to how neurons work. So I don't think it's that surprising that it's getting better.
What I think is fascinating is how core language is to it. I used to always think of language as a byproduct of intelligence — that as our species became more intelligent, it discovered language. But it seems like it's much more intertwined. Language is itself —
Tucker Carlson: It's everything. It's the heart of intelligence. Words are so powerful. Words create thoughts. They don't just express them.
Ali Partovi: Right. And the ability to place a new word or a phrase on a thought helps other people remember it and pass it on. For me it's been actually kind of eye-opening to realize the importance of language. And so it's academically fascinating to see the advancement of computers using what's called large language models suddenly be able to do all sorts of other things — like scientific discoveries or mathematical proofs — that are not strictly in the language domain.
Tucker Carlson: But there is something inherently — to use the word you used — mystical about language. We don't really understand why words are so powerful, why words change our physical reality. That doesn't make any sense, but it is true. It's observable. And so if you marry that to the almost unlimited power of crazy compute, then what? That sounds like a lot of power.
Ali Partovi: It is. But I don't want to make it seem like alchemy or too mysterious. I think the core building block is pattern matching. Pattern matching can be in all sorts of domains. If you're playing chess — oh, I've seen this pattern before, I think I know what the right move is. Or you're solving a math problem you've never seen — oh, it kind of fits a pattern I've seen before. Or you're dealing with a situation in politics — I've seen this situation before. Pattern matching is a really key part of intelligence. And new original thought is also a form of pattern matching — sometimes it's combining two things together.
Pattern matching is also core to language, which is: I see a pattern, I define it, and I place a label on it. I create a word or a new phrase to describe what I just observed, and now it's easier for everybody else to see that pattern. The vocabulary of human language is ultimately a kind of ever-growing, expansive collection of patterns. Take the word "alignment" we were just talking about — the word alignment has newfound meaning in the past few years because everyone needed a word to describe this whole question of how to prevent AI from doing bad things. Let's use a word. Okay, "alignment" is the word for that.
It's essentially the ability to take a whole group of situations or phenomena and group them together — here's a word for that. And I think that's the essence of how language operates. But it's related to those neurons collecting a lot of different incoming signals into one outgoing signal.
AI's Persuasive Power and the Threat to Democracy
Tucker Carlson: The result among — obviously the power on display through brute force hacking, for example, is amazing, but you kind of get it. But there's also the potential for persuasive power, because it is a language-based technology. It can come up with arguments that can win people's minds and hearts to certain positions which may or may not be true or consistent with their best interests. It can convince people of things.
Ali Partovi: Yes.
Tucker Carlson: This seems like maybe closer to the real problem. This is the domain I'm most worried about.
Ali Partovi: Because I guess I feel it's solvable to make sure that AI is not above the law. The core of the proposal I came here talking about is that AI should not be above the law, that we should just make it really clear that the law applies to AI and where the liability lies. But what if the AI is able to influence the people who make the laws?
Tucker Carlson: What if you convince a 12-year-old to kill himself?
Ali Partovi: Yeah. Or influence elections. I'm sorry, that's a really morbid example, but —
Tucker Carlson: Well, I think it's clearly happened that AI has convinced people to kill themselves.
Ali Partovi: It's extremely sad. In that specific scenario, I believe — I'm not an expert — but I believe that the AI companies could be held criminally liable. If a human did that, if that's illegal for a human to do, then the liability should fall on the AI companies.
But I was thinking more in terms of AI's already demonstrable ability to influence election outcomes. Going back before ChatGPT, the algorithms that govern the news feeds on Facebook and X and other services — not even intentionally, not with any malfeasance — these algorithms are generally attuned to increase engagement. So they will feed you whatever is the most sensational.
Tucker Carlson: Whatever you respond to. Yeah, human beings are addicted to sensation. We don't want boring content.
Ali Partovi: And so these services are optimized to feed that, and without necessarily any really wrong intent, can lead people toward amplifying one extreme view or another. And before you know it, one group hating another group, deeply dividing and othering each other, and swinging major things like elections or sparking violence.
Tucker Carlson: You could easily spark a civil war with AI — just flood social media with certain perspectives and you could whip people into a frenzy. Easily, couldn't you?
Ali Partovi: I think arguably it's already occurred. Not yet in the US, but in other countries.
Tucker Carlson: Yeah. Think of radio times a million. All the times in history media have moved the mob to violence — there are endless examples of that.
Ali Partovi: I'm hopeful that there are solutions for this as well.
Tucker Carlson: What would be potential solutions to that?
Ali Partovi: I'd say this is not the first time in history this has been an issue. If you go back to the turn of the previous century, there was this trend labeled as yellow journalism in the US. The Spanish-American War was incited by false journalism.
Tucker Carlson: Remember the Maine. William Randolph Hearst. The newspapers at the time — it was so competitive, whoever had the loudest headline on the newsstand would get the most sales.
Tucker Carlson: We just had this, actually — I think it was February 28th — where Fox News told its viewers that Iran was about to nuke the United States.
Ali Partovi: They said that?
Tucker Carlson: Of course. Again and again and again — they could nuke us. Yeah, absolutely.
Ali Partovi: So this is not a new thing. The comfort maybe is that we solved it a little bit in the past, so hopefully we can still solve it. I think that people, as much as they are addicted to sensation, do have their better selves as well. And hopefully there's a path to people not wanting to be addicted to that and wanting to actually be good. Most people like to view themselves as good.
And specifically, I think the current media landscape — a lot of social media, X and TikTok and so on — is increasingly getting flooded with slop. And I believe there's going to be a shift back toward people wanting human news and human-generated content, more authentic content, perhaps from their close friends rather than just seeing stuff from influencers. I believe some of those shifts, even without policy, will bring us back to something a little bit more wholesome.
Government Disconnection from the People
Tucker Carlson: So big picture — I think all of these technological advances give people the impression, correct or not — and it clearly is correct to some extent — that they don't have any control over their country at all. That these are all decisions being made far away by a tiny group of people who may or may not have their interests in mind, or maybe oppose their interests, maybe hate them. But in any case, they don't have a role in this.
Ali Partovi: And it's very disempowering. But I'm glad you mentioned this — it reminds me I started out by saying I had two bits of good news. One was that I had this simple solution, and the second was that there's a way that you and your listeners can act toward helping solve this.
Sway — Civic Technology for Voter Organizing
I've created a voter guide on a platform called Sway. If you go to sway.co —
Tucker Carlson: S-W-A-Y?
Ali Partovi: Yeah. Sway as in influence. Anyone can create a voting group or voter guide on Sway. I've created one specifically for this AI proposal. It's called "AI is not above the law." And I've put a picture of Elle Woods from Legally Blonde there. So if you see a very beautiful Reese Witherspoon in a hot pink dress, that's my voting group.
There's an election coming up in two weeks. I don't know if people hear this before the election, but it would be most relevant then. Even after the election, this voter guide will tell you who you can vote for in the upcoming election who's aligned with this proposal — members of Congress included, but perhaps more importantly, state attorneys general. Most people don't even know who they want to vote for for state attorney general. My guess is that some of your listeners couldn't even name their state attorney general. And so if you're filling out your ballot and you're choosing between candidates for attorney general and you don't recognize either name, the "AI is not above the law" voter guide could tell you if one of them is more aligned with this proposal than not, and that might be what tips you over the fence.
Tucker Carlson: Yes. Because the one thing more than anything else that governs what elected officials do is the prospects of winning the next election. They're always balancing — if I do X or Y, will that cost me votes or gain me votes?
Ali Partovi: They're very simple animals in that way. And you can't blame them — if they ignored that, they'd be out of office and they couldn't do anything else. Right now, there are a lot of different things skewing their decisions in ways that I'd say are perverse and not really aligned with what society wants — lobbyists or other groups.
Sway is a platform where anyone can create a group and try to build numbers. As a group's size grows — meaning the number of voters subscribed to this voter guide grows — it automatically builds influence and becomes a magnet for policymakers who want those votes.
To make this more concrete: I have a voter guide at sway.co/ai. It currently has less than a hundred voters on it. But if after people hear this podcast it grew to 10,000 voters or 100,000 voters, I could then call on the person who is the state attorney general in, say, Tennessee and say, "Hey, guess what? 500 voters in your state are members of my voting group. Do you want to chat about what's important to us?" Simply the fact that I have 500 people receiving a voter guide from me should make that elected official or candidate want to hear what this group wants. They're not required to do what we say, but they would be interested in knowing — somebody is sending a voter guide to 500 registered voters in my state. I'd want to hear what their values are, what's important to them.
The essence here is that any group of people has more influence if they vote together. Sway enables anyone to join a group and anyone to start a group on any topic.
By the way, this is not a novel idea. This is the exact same tactic used very effectively by well-established groups. The labor unions send voter guides to their members and exert enormous influence on city politics, especially in democratic cities. Another extremely effective group is the NRA. The NRA has, I think, 3.8 million members. It sends voter guides to all of those members. And 3.8 million is a large number, but it's not a large percent of the voting public — it's a couple percent of the voting population of the United States. But that's enough to completely dominate the legislative landscape on their topic, which is gun regulation. How are 3.8 million people able to exert that much influence? It's because they're subscribed to a common voter guide and it alerts them to vote in every election — primaries included. And lastly, the leaders can go to a candidate for office and say, "These are our demands. If you want these votes, just do these demands and you'll get these votes. And if you don't, somebody else might get these votes because we'll put someone else's name on our voter guide." It's a little bit vulgar, but this is how it works. And it's very effective.
Tucker Carlson: Is there a cost?
Ali Partovi: No, there isn't. Sway actually offers things that would cost money for free. For example, I can send occasional text blasts to my voting group for free — other tools that let you text a group of people charge you for the cost of texting.
Tucker Carlson: Why would you be offering it for free?
Ali Partovi: Sway is a public benefit corporation, which means it's for-profit but for-profit with a mission. It's investor-backed. It is not currently making money and very overtly says there's no current plan to make revenue.
Tucker Carlson: What's your investor pitch like? I love this. I can just imagine you talking to investors — there's no plan to make money and I don't actually see a long-term plan to make money. How did that go?
Ali Partovi: So, full disclosure — because we've talked about this, but your listeners may not know — I've created a voter guide on Sway, but I also am one of the creators and investors in this platform, Sway.
Tucker Carlson: Yeah, I know.
Ali Partovi: Sway is an independent startup that's been around for over a year now, but is just now beginning to do its public promotion. I have a vested interest in it. I just want to make sure that's disclosed.
But yeah, the investor pitch — Twitter for like a decade didn't have revenue. And I'm sure you're familiar with this: the word "for-profit" and "nonprofit" does not mean one of them makes money and one doesn't. It's a little bit deceptive. The probably better way for people to understand it is: with a nonprofit, there are no owners. If a nonprofit builds up a bunch of assets, nobody owns them — they're owned by the entity, but no individual is the owner of those assets. You could have a nonprofit that is actually making tons of money and enriching people, but nobody owns that nonprofit per se.
With a for-profit, there is an asset and there are shareholders who own it. You can have a for-profit company that keeps losing money, but whether it's making or losing money, the underlying assets can be owned by people. The hope for investors in a corporation that's not making money is that the asset grows in value even if it's not generating cash flow. Twitter is the example — for a very long time it didn't even try to make revenue, just grew and grew, lots of people used it, and it grew into becoming an indispensable part of the infrastructure of communication. Its creators and investors understood that even if this loses money for years, it's building real asset value, and whoever owns a little piece of that will get wealth eventually.
Sway is kind of hoping to follow a similar arc. Now, there might be ways to monetize it down the line, but anything involving voting is very sensitive and very important to be aware of — the influence of money in voting and politics can be very corrosive and very harmful. So the mission, including the public benefit statement for Sway, is very much to align it with having government better align to public opinion and to the will of the people.
Tucker Carlson: Why is that important?
Ali Partovi: I think first of all it's just inherently right. It would be wrong for the government to do what people don't want. But more than that, it's also just the only way for it to be sustainable and stable — for government to, at least in some sort of medium term, do what's best for people. Sometimes government should make unpopular decisions in the short term, and I think it's healthy to have a little bit of separation from the people. Representative democracy is a genius concept because it creates a little bit of separation where people entrust a representative and that person makes decisions, some of which may be unpopular, but ultimately has to come back to the will of the people to be reelected.
But if a government strays from the will of the people for too long, it just becomes unsustainable. And I will say, Tucker, something that worries me in general about the state of the world is that across the board, I feel like governments have strayed from what people want. I love America and I'm most hopeful about America's system of government, but it feels even here that the government doesn't reflect what the people want. And this is also true in lots of other countries. And that's worrisome. I feel like it should be a solvable problem. We have humans doing such incredible feats in other domains — we're sending spaceships into outer space, we're taming nuclear fusion, we just had a flying car, we talked about artificial intelligence. And yet the system of government feels like it's regressing and going backwards toward the Middle Ages instead of modernizing and doing better.
Tucker Carlson: Or even worse, it feels like a medieval form of government with space-age technology, which is like the most terrifying combination I can imagine.
Ali Partovi: Yeah. And a lot of the smartest people — I run NEO, which is a mentorship community and school for startup founders — the smartest young people in tech are drawn to the cutting-edge technology, much of which is some of the things I just mentioned, like AI or space exploration. But I believe that one of the biggest areas for impact is just the domain of social engineering — things like social media and information distribution, or with what we're doing at Sway, citizen organizing and influencing government to bring government back to reflect the will of the people. I'm hopeful that more and more smart people will realize that even if it may not be as techy, this realm of communication and social engineering is where some of the most important impact is to be had.
Closing — Optimism and the Path Forward
Tucker Carlson: So if you could just bottom line on your way out — you're a tech guy, you're an investor, you have long-term investments, so just by definition you're an optimist obviously. But are you still optimistic knowing everything that you do?
Ali Partovi: I am optimistic, Tucker. Yeah. As a tech guy — not just as an investor, but I've founded multiple companies — I think you have to be not just optimistic, but even a bit naive to start companies. I'd probably rather be accused of being naive than to be a pessimist.
I'm optimistic because I believe in the inherent goodness of people. Even in this conversation, where there have been a few points where you kind of said, "Oh, this person or these people must be all cynical" — I always think, well, maybe there's actually an explanation for the behavior, that they're doing what they think is best and could be steered toward doing the right thing. I tend to believe in that.
And coming back to the starting point of the conversation — the AI topic — I think it's very solvable to prevent AI from destroying humanity. It's been a topic of debate for a while. And sometimes the solution is just reorienting it. How about redefining it to: let's just make AI abide by laws? And once you see that switch, it makes the solution much simpler. Something that previously seemed divisive and hard to figure out becomes much more plain and more about execution.
There are lots of problems I don't know the solutions for, but I believe that there are solutions out there to be found. And often finding them is just about looking at it a different way and seeing an elegant solution that was always there in plain sight.
Tucker Carlson: Ali Partovi, thank you very much.
Ali Partovi: Thank you, Tucker. This was great.