Wipe out a ‘civilization’? Minor stuff compared to what just happened in AI

Many of us have heard of our president lately.whole civilizationA technology company in Silicon Valley was warning that it could be accidentally disrupted without much notice. All civilization as we know it.

San Francisco tech company Anthropic said in a statement on Tuesday. wasn’t Claude AI is launching a new version of its superbrain – because it’s so powerful it’s capable of breaking into almost any computer system, no matter how secure, in a matter of days, if not hours.

“The consequences for economies, public safety and national security could be serious,” Anthropic said. he said in a statement.

Concern about artificial intelligence is nothing new. We worry about artificial intelligence taking over, about toys that look too real for our children, about our every move being under mass surveillance. But Anthropic’s warning about its own product is bigger than these individual problems. It’s a call from inside the house that disaster is hiding just around the corner. I know it sounds extremely dire and exaggerated. But here’s the thing; It’s not like that.

Antropik, you may remember, was the company founded by US “War” Secretary Pete Hegseth. is fighting because he didn’t want Claude to go into battle unsupervised and maybe do something like that. accidentally bombing little girls at school.

Now, that company revealed this cooling warning: The current Claude that caused this mess is outdated and surprisingly less powerful than the new one; He’s trying so hard not to let loose – though this is the new Claude, Called Claude Mythos PreviewHe’s already escaped on his own at least once. I’ll tell you more about this in a moment; There is only so much existential fear a person can handle.

“We should all be worried,” Roman Yampolskiy told me about this latest advance in a technology that is sure to change the course of humanity. He is the country’s leading artificial intelligence security researchersand professor at the University of Louisville in Kentucky.

“We are about to create a general superintelligence, and this threatens humanity as a whole,” Yampolskiy said.

“Everything else is irrelevant,” he added, before suggesting that I stop calling myself an idiot for not understanding the tech-heavy parts of this discussion. He assured me that my simple approach was “a reasonable way to explain this.”

Here you go.

This is the “really smart computer geniuses could abuse this” scenario or “everyone will be unemployedscenario, or even the “may accidentally bomb children” scenario, which is a really terrible scenario.

This is the “your teenage son can use this to hack into the local school district system, change his classroom with very little knowledge, and accidentally destroy the California power grid” scenario.

Or maybe a country that doesn’t like us — a few come to mind — could drain every U.S. citizen’s bank account, while also unlocking the automatic locks of prison cells, shutting down our sewage plants, and taking over air control systems. Or maybe Claude Mythos is doing it alone.

For example, Antropik said Claude Mythos said he found a flaw that had gone unnoticed for 17 years in a popular operating system he tested, used by thousands of companies, including Netflix and Sony. He then figured out how to exploit the flaw on his own—without human guidance or assistance—to take control of any server running the operating system, using any computer anywhere in the world.

I’m rambling here, but if almost no security system is secure, the possibilities for social, financial, and general chaos are truly endless. And honestly, any security expert will tell you that some of America’s biggest weak spots when it comes to cybersecurity are local and state governments, because oddly enough, top experts don’t work five-figure jobs in cities on the Great Plains.

Based on your own tests, Anthropic predictions It can find “more than a thousand critical severity vulnerabilities and thousands more high severity vulnerabilities.”

This means Claude Mythos is putting our infrastructure at risk everywhere – because so much is connected through backdoor ways that most of us never consider that it only takes a weak system to open the door to hundreds of others. But it is nearly impossible to protect and remediate all these systems quickly and robustly enough against this type of AI.

Yampolskiy said it was just a cybersecurity risk. An AI with Claude Mythos’ abilities could be used to advance in many more areas.

“We’re seeing the same thing happen with synthetic biology. We’ll see the same thing with chemical weapons, probably something new in terms of weapons of mass destruction,” he said.

To Anthropic’s credit, it did provide caution regarding its creation and created something of a game plan, if not a solution. Glass Wing ProjectI suspect the name because no matter how bad the situation is, we will make it look like a thriller with an exciting ending.

Glass Wing Project It would be better if the project was called Headstart, because that’s what it is. Before releasing Mythos into the wild, Anthropic is releasing it to about 40 tech companies, including Apple, Google, and Nvidia, to see if they can collectively fix all the vulnerabilities they find before giving the general public a chance to reach them. It’s like the killer in the movies giving the victim 15 seconds to escape.

So, I’m going to take 15 seconds and hope these things come true. But as Anthropic said in a statement, “the work of defending the world’s cyber infrastructure could take years; edge AI capabilities are likely to advance significantly in the next few months. We need to act now to help cyber defenders step up.”

So do we really have 15 seconds? One of Claude Mythos’ scouts recently posted on social media that he had lunch in a park. Mythos emailed him – even though he wasn’t supposed to have access to the internet. The researchers tasked Mythos with trying to break out of the disconnected “sandbox,” and it succeeded.

This Another problem with Mythos and other AI — they rarely do what we expect and find sneaky ways around the rules. Nearly every AI superbrain ever created has been shown to lie, cheat, and generally behave in disturbing and unethical ways given the right conditions.

Even Claude, who has been called one of the most ethical AI superbrains out there, engages in bad behavior. Anthropic boasts It’s the “most well-aligned model” ever made—it’s a model that speaks to technology to follow human values ​​and intentions, but also acknowledges that it “probably poses the greatest risk to alignment,” which is something that technology speaks to, perhaps it isn’t.

So, at least for now, being the most ethical AI superbrain is a bit like being the most ethical serial killer. Run people, run.

Thank you again Antropik (and his Chief Executive Officer Dario AmodeiI’d be willing to bet that because he doesn’t plunge us into global chaos without warning, he frequently warns about the dangers of what he’s creating, because some other corporation might have dumped their super AI on society and let the destruction fall where it should. There will no doubt be other AI minds as talented as Mythos, and soon Anthropic has taken first place with this level of talent, but only 15 seconds ahead of its rivals.

But the idea that the tech industry will or should solve these problems on its own is a massive abdication of duty and common sense by governments large and small to protect their people. This is not a contest for dominance as President Trump has described it. It is a race to protect ourselves from ourselves and from the majority of the super-rich titans of industry who continually put business and commerce ahead of the social good.

We’re down to the last 15 seconds before AI changes everything. We either demand control and regulation now, or let tech companies determine the fate of the world.