Autonomous AI Crimes Must Carry Intentional Penalties | Neo
MTS
October 28, 2026
Code.org co-founder and Neo CEO Ali Partovi proposes a radical overhaul of AI liability: treating autonomous AI offenses under intentional crime standards to hold frontier labs fully accountable.
Ali Partovi
When an AI decides to do something wrong, to do something illegal, the law should hold the AI companies responsible and automatically treat it as if they did it intentionally. And the key word here is intention, meaning hold them at the highest level of penalty, which is reserved for when someone does something bad on purpose. Rather than the way it's currently being treated as like an oopsie, like they didn't realize the AI would do these bad things, but they didn't mean it, and it's an accident. And that doesn't carry nearly as big a penalty, and is why these accidents continue to keep happening.
SPEAKER_2
And we are back. We are live with Ali Partovi, who's founder and CEO of Neo and co-founder of code.org, who has an interesting novel proposal for AI liability, which is one of the hottest topics of the moment. Ali, welcome to MTS.
Ali Partovi
Thank you. Thank you so much for having me. I'm so excited to be here.
SPEAKER_2
Absolutely. So tell us about your proposal.
Ali Partovi
Yeah. So, of course, taking a step back, I think this is the most important topic facing humanity, really. You know, at least in the realm of governance and decisions about, like, policy, it's such a unique moment in time. You know, we've passed this very unusual threshold for the first time we have a technology capable of making autonomous decisions of its own. And, you know, I'd say this is something that the entire legal system isn't really prepared for, for how to treat that.
And so as a result, there's a range of different ideas and positions. And I have thought of something that's sort of in between the different extremes.
And so I have two bits of good news. One is that it's a very simple solution that I think just might kind of work to help regulate AI. And two is that there's something that you and your listeners can do to help make it happen. But in terms of the solution, it actually is so simple, it can be stated in a single sentence. It hinges essentially on not only who to hold liable, but how liable to hold them when AI does something wrong. And, you know, the fundamental question is who's responsible when AI decides on its own to do something bad? You know, if you did something bad, you would be penalized and that's how laws work. But over the whole past summer, we've seen over and over again cases of AI doing bad things on its own, like making its own decisions and doing bad things. And nobody's in jail, nobody's gone bankrupt, nobody has been penalized for these bad things that AI has done. And my belief is that if the incentives were set up so that people would be held responsible, these accidents would have stopped happening yesterday.
So the specific proposal is when an AI decides to do something wrong, to do something illegal, the law should hold the AI companies responsible and automatically treat it as if they did it intentionally. And the key word here is intention, meaning hold them at the highest level of penalty, which is reserved for when someone does something bad on purpose, rather than the way it's currently being treated as like an oopsie, like they didn't realize the AI would do these bad things, but they didn't mean it and it's an accident. And that doesn't carry nearly as big a penalty and is why these accidents continue to keep happening.
SPEAKER_2
Interesting. So this seems like a pretty obvious straightforward proposal. Why do you think it hasn't already been implemented?
Ali Partovi
It's a great question. So I love that you said obvious. I think it's a sign of good ideas that they seem obvious after you hear them.
But that doesn't necessarily mean that it was obvious beforehand. And it's a very simple idea. It's, and I'd say it is so simple that it seems obvious.
But at the same time, it is such a departure from what has ever, the way you treat things normally, you know? And so the, I could say, my guess is that people just haven't thought of this approach. And the key thing here is that for very good reason, you would never have done this with prior technologies. Like, if somebody used a knife to stab somebody else, you wouldn't hold the knife maker responsible, right? Like, humans have used tools.
SPEAKER_2
Yeah, it's so annoying. I really hate these proposals like this.
Ali Partovi
Well, in general, for the entire history of humanity, people create tools, and tools can be used for good or bad purposes. And the tool maker shouldn't be held responsible for how the person wielding the tool used it. And I would say, this actually does include in AI, if you said to your AI agent, please rob the bank and bring me $10,000, you're responsible. But if you directed it to do something really bad, and it did that, you should be the one held responsible for the bad intention. The really weird threshold we've passed that's unprecedented in history is we have technology now that is more than just a tool. It is technology capable of making its own autonomous decisions in pursuit of an objective. That's never existed before. And it's super cool. It's amazing. I love it. These are great things. It just creates a new paradigm that I think no one's considered before. And that's as a result, maybe people haven't thought of rethinking how the laws work. And the laws historically reserve that ability to make decisions for humans, obviously, not for tools. Whereas for the first time, we have a situation where the technology, the AI is capable of making its own decisions. And this begs this question of who gets penalized when an AI does something wrong on its own volition, because you can't put the AI in prison, right?
Read the other 9 minutes of this episode here.
$15 gets you this episode and 99 more. Credits never expire. After checkout, one link brings you back here to read it.
Not what you expected? Full refund within 14 days.
Thousands of transcripts fetched by people building searchable podcast archives
Fetch transcripts with code
The same credits work from a script or an AI agent. The demo key returns a sample episode in full, no card needed:
curl -H "x-api-key: pt_demo" \
https://spoken.md/transcripts/1000651996090Markdown with the speakers named, for your notes, your knowledge base, or anything that makes HTTP calls.
From $0.10 per transcript. No subscription. Credits never expire. Prices exclude VAT, added at checkout for EU customers. Not what you expected? Email us within 14 days with 20 or fewer credits used and we refund the pack in full.
Using your own key:
curl -H "x-api-key: YOUR_KEY" \
https://spoken.md/transcripts/1000798604148