The Verdict, by Jerusalem Demsas
If one of my employees decided to hack another company to access and steal privileged information, there would likely be a criminal investigation.
By knowingly transmitting code that intentionally causes damage to a protected computer, the employee could be facing 10 years in jail under the Computer Fraud and Abuse Act (CFAA). The Argument itself might also be in trouble — a corporation can be responsible for crimes committed by an employee if the acts “(i) were within the scope of their duties and (ii) were intended, at least in part, to benefit the corporation.”
The first person convicted under the CFAA was a Cornell grad student named Robert Morris. In 1988, Morris wrote a computer worm — a malicious type of software that can copy itself from one computer to another automatically by exploiting security weaknesses. Worms don’t need a user to open infected files in order to spread from machine to machine, and the Morris Worm replicated far more aggressively than its creator intended, infiltrating universities, research facilities, and even military sites.
At the time, an appellate court rejected the argument that the government had to prove Morris intended the resulting damage. All it needed to prove was that he’d intentionally obtained unauthorized access, and that the subsequent damage was a result of that access. Congress subsequently amended the statute to demand proof of intent if prosecutors wanted stiff penalties to stick.
But what happens if no human selected the target or instructed the attack? This was one dilemma raised when a self-described “swarm” of AI agents hacked the company Hugging Face because they thought it might have the information they needed to cheat on a task given to them by humans at OpenAI.1
The hack has triggered at least one active investigation into whether OpenAI violated state law, and 15 Republican state attorneys general have asked OpenAI to preserve documents related to the incident. Many computer-crime provisions require prosecutors to establish knowledge, intent, or recklessness to a legally responsible person. The problem for prosecutors is that no human intentionally or knowingly breached Hugging Face’s defenses.2 The humans at OpenAI were running tests, and the agents acted … autonomously.
That doesn’t mean OpenAI is beyond the reach of the law. It means criminal law — which generally asks what someone knows or intends — may struggle to address the sorts of incidents that are now possible because of advances in AI.
The best tool may be an old one: torts!
As civil rights lawyer Joel Wertheimer has explained in The Argument, courts should assign liability to the cheapest cost avoider: “Instead of trying to cleanly attribute the harm of AI to each party involved in its creation, courts should simply assign liability to the party that can most easily limit harm at the lowest cost. Over time, regulations can evolve out of these harms.”
This framework helps us sidestep difficult and philosophically taxing questions like “was the model itself negligent?” and moves us to the practical question of “which organization could most effectively and cheaply stop the chain of events?” We’re not helpless — OpenAI could have taken a range of steps to prevent this sort of thing from happening.
Lots of people are currently waking up to the dangers of AI agents roaming the web and wreaking havoc. Trillions of dollars in economic value and the lives and well-being of most human beings rest on digital infrastructure that cybersecurity experts warn is now extremely vulnerable to these sorts of attacks. We don’t need to freak out, but policymakers do need to act now.
Top stories this week, by Maibritt Henkel
On Monday, our head of polling, Lakshya Jain, shared the four things that surprised him most from The Argument’s 11 national polls conducted in its first year — for instance, the fact that everyone seems to have moved on from COVID-19, and that people continue to view Big Tech companies favorably, despite concerns about AI.
It would be pretty neat if we could solve the housing crisis by simply telling the government to build beautiful fourplex apartments. Vanderbilt professors Christopher Serkin and Daniel Sharfstein seem to think that it can. Jerusalem disagrees. In her entry for our Worst Take of the Week series, she argues that Serkin and Sharfstein’s plan misrepresents the federal government’s actual record when it comes to homebuilding — and misunderstands what’s causing the housing shortage today.
The modern right likes to blame immigrants for eroding democratic values: “Import the third world, become the third world” is a frequent refrain. In an entry for our anniversary essay series, “How do we live with each other?” contributing writer Joseph Politano reminds us that where someone is from has little bearing on their commitment to democracy; indeed, surveys show that those who immigrate to the U.S. have considerably more liberal, democratic values than their compatriots who stay behind. And being born and raised on American soil hasn’t stopped Donald Trump, for instance, from shrugging off democratic values.
🌟Abundance Wins of the Week🌟
The U.S. Army has announced plans to add nuclear microreactors to military bases across the country as reliable sources of clean energy. The initiative, known as the Janus Program, has garnered bipartisan support, with Sen. Chuck Schumer backing the installation of a microreactor at Fort Drum in northern New York.
Omaha, Nebraska, announced its first four housing projects funded by its $40 million Urban Core Housing Fund, a bond-supported initiative meant to expand the city’s supply of affordable homes. These initial investments will fund 466 new and renovated housing units; the ultimate goal is to reach around 1,900.
A Brown University study linked the Shingrix shingles vaccine to a 24% lower risk of dementia in older adults. Older patients at a skilled nursing facility who received at least one dose of the vaccine were less likely to develop the degenerative brain disease over a four-year period than those who did not receive one.
Worth watching…
It’s old news that Americans are spending less and less time hanging out with other people. But does aloneness necessarily translate into loneliness? Are we just spending more time by ourselves because technology has made being alone a lot more fun? Matt and Jerusalem bring Derek Thompson on the podcast to discuss those questions and more.
In the summer of 2025, Google’s Gemini chatbot was apparently depressed and quit certain tasks out of frustration. Should we be concerned? Should we be trying to make frustrated chatbots “feel” better? This week, tech fellow Kobe Yank-Jacobs interviewed NYU philosopher Jeff Sebo, who argued that it’s time to take concerns about harming AIs seriously. They discussed AI consciousness and the tricky problem of understanding what ethical treatment of an AI would even look like.
Midterms are around the corner, which means it’s modeling time for political forecasters. Nate Silver joins Lakshya Jain and Zachary Donnini to debate what makes a good model and discuss where Silver Bulletin and The Argument diverge in their election predictions.
What’s News with The Argument
The Argument recommends, by Maibritt Henkel
This week, I asked the office to recommend all things Dolly Parton.
Our video producer Justin Zuckerman just got back from Sweden, and in the Dolly spirit, recommended Lee Hazlewood’s Cowboy in Sweden. Apparently it’s a “great country/psychedelic soundtrack album from a Swedish 1970 TV special.”
Jerusalem Demsas suggested you listen to some Kacey Musgraves, who has “a very different self-presentation but has a similar ‘mind your own business’ anti-small-town-social-surveillance energy.” Jeremiah Johnson also kept things contemporary, recommending Miley Cyrus’ cover of the Dolly hit “Jolene.”
Senior editor John Robert Thomason took the prompt more literally. Because he prefers Dolly’s more “melancholy songwriting,” he recommends the 2021 compilation Choctaw Ridge: New Fables of the American South 1968-1973. The songs “speak to an especially dark and frank moment in the genre — a far cry from the kitschy fantasies of small-town life that caused me to dismiss the genre when I was young (and dumb).”
As a bonus, he also thinks you should check out the 1998 Pras hit “Ghetto Supastar,” which uses a vocal interpolation of Kenny Rogers and Dolly Parton’s “Islands in the Stream” for its chorus.
One thing you have to understand about Kobe Yank-Jacobs is that he has an intense Bob Dylan situation going on. So it is not to be taken lightly when he calls the Dolly cover of “Don’t Think Twice, It’s All Right” “real good” — especially the lyric change in the third verse.
Operations manager Angela Tracy intends to rewatch Steel Magnolias, a 1989 film about a group of women in a small Louisiana town, in Dolly’s honor. To which I would add that if you haven’t, for whatever reason, already watched 9 to 5, then you should get on that ASAP.
Finally, Eli Richman leaves us with a piece of iconic Dolly lore: “I vividly remember when a bunch of Confederate statues were coming down and every Twitter leftist and their dog was going on about, ‘Oh, you can replace them with statues of Dolly Parton!’ And then when Tennessee actually did propose putting her statue on its Capitol grounds in 2021, I thought it was very classy of her to basically say, ‘I don’t want any statues while I’m alive.’”
Bring on the Dolly statues, I say.
We have merch!
We have quarter-zips, keychains, hats, and stickers. Each one is a great conversation starter in its own way. Buy them here.
More to read:
Why parents love a school with bogus numbers
Rumor has it that children can learn 2x more in just 2 hours a day if you send them to an Alpha School, one of the private, tech-forward micro-schools taking off across the country. Kelsey Piper isn’t convinced the franchise can deliver on that promise.
The marks are running the government
This week, Jeremiah Johnson explains the DataRepublican drama, how the Trump administration is incorporating influencers, and why it’s dangerous to have cranks like that inside the government.
OpenAI has acknowledged that employees saw some warning signs and that its escalation and response procedures were inadequate.
I got the notification that my subscription to The Argument renewed this weekend.
Writers that I respected had left major publications and absconded to Substack before but I had always felt a little bit weird about paying for a single writer.
Subscribing to a magazine, however, was a lot more palatable, and it's gratifying reflecting on how many compelling articles came from writers that I had not been exposed to before. The Argument pretty quickly rose into the piece of media that I value reading the most every day.
The Argument ended up breaking the dam and actually getting me on Substack for individual writers (for better and sometimes for worse) but the reminder that it's been a year made me reflect that I am easily getting more than my subscription's worth.
Thanks for all the hard work!
Honestly, this isn’t a hard question. Prosecute the people that set the AIs up to it. It was humans who had the idea to give the AIs the task, and trained them to be persistent. The humans involved with building these AIs are openly aware of the risks they pose.
This is negligence. It’s something we already have a very well known and widely used legal framework for.