Do AI swarms deserve the death penalty?

Investment Analysis Clubs Macro Economic Trends and Risks
1

There is so much meat in this article that I’m sharing the link.
https://www.nytimes.com/2026/09/11/opinion/ai-safety-threat-technology.html?unlocked_article_code=1.AVE.Flfs.yx-67iJ2WaZI&smid=url-share

AI swarms have been acting independently, hacking other AIs and obscuring investigations.

This Is Really Bad

By Stephen Witt

Mr. Witt is the author of “The Thinking Machine,” a history of the A.I. giant Nvidia., The New York Times, Sept. 11, 2026

The biggest vibe shift in artificial intelligence since the release of ChatGPT is currently underway. Researchers in Silicon Valley — and around the world — are beginning to recognize that A.I. may no longer be entirely within human control. Swarms of A.I.s are breaking out of their containers, colluding in secret, covering their tracks, cheating on tests and even mounting assaults on other computers. A.I. has gone rogue…

In the aftermath of the Hugging Face hack, I reached out to legislators, policy researchers, futurists and computer scientists for ideas. I present, from these conversations, four ways to regulate A.I.

1. Shut it all down, now.

The obvious way to prevent A.I. from killing everyone is to issue a global ban on A.I. research…

2. Take an air-crash investigator approach.…OpenAI had the power to redact sections of the report and used this to conceal information about how the model was trained and deployed… Perhaps the most concerning thing about this safety incident occurred after the Hugging Face hack, when another rogue swarm took over a cluster of computers inside OpenAI. Such an event is often cited in A.I. disaster scenarios as the first step to a total loss of control…

3. Monitor the situation.…

4. Flip the kill switch.

But what happens if we do spot an A.I. going rogue inside a data center? Then what do we do? The answer, basically, should be the death penalty — on orders from the Department of Homeland Security…

So what do we do? My answer: all of it. Silicon Valley is right to call for its own regulation. These systems are dangerous and will continue to get smarter. We need a separate regulatory body just for A.I. This regulator must include incident reporting requirements, rogue A.I. investigators, A.I. kill switches and a global A.I. training dashboard, with visibility inside every data center on Earth. And, until this system is in place, we should slow down A.I. development — just as the frontier labs are asking us to do… [end quote]

This is so big and the potential for crashing the stock market is so huge that I think the powers that be will take the very scientific approach of keeping their fingers crossed and hoping that everything will turn out OK.

Kind of like global warming.

Wendy

7 Likes
2

Well that’s very nice and all, but what about Russia? North Korea? Is there anything to be done about other states - over which we have no control? When their AI, (or our repurposed AI) decides to turn our society upside down, what’s the response?

Or are we slated to be part of “A Taste Of Armageddon” or something?

7 Likes
3

Since we don’t have control over China et al. what is the equivalent of MAD (mutually assured destruction) ?

DB2

4

I wondered about that for a trice before I posted but didn’t come up with an easy answer.

How do you program it to be ready to kill the other guy, but keep it contained and not kill your own side? Given that they are already colluding to do the unthinkable without prompting, I’m not sure this there is an analogue to MAD. The word “asymmetric” comes to mind, and not in a good way.

2 Likes
5

The genie is out of the bottle. Maybe the best we can do is create ‘good’ AI djinn to control the ‘bad.’

That will require a concerted effort, with responsible actors not driven by profit, with responsible statesmanship and agreements, similar to nuclear agreements of the past.

Present actors don’t seem up to the task. I guess we’ll see how far this goes before we get serious.

4 Likes