Mythos konnte Tausende bisher unbekannte („Zero-Day“) Schwachstellen in wichtigen Betriebssystemen und Anwendungen identifizieren. Darüber hinaus kann es funktionierende Exploits generieren und nicht nur theoretische Fehler identifizieren. Als ob das noch nicht schlimm genug wäre, kann dies auf einem Niveau erreicht werden, das mit dem Niveau menschlicher Top-Experten vergleichbar ist oder dieses sogar übertrifft.
Banken und Finanzinfrastruktur sind besonders gefährdet. Sie sind a) stark miteinander verbunden. b) Abhängig von Legacy-Systemen (häufig mit versteckten Schwachstellen) und c) Systemrelevant (Ausfälle können sich global ausbreiten).
Die USA spielen "Verpiss dich und finde es heraus" Bei so vielen Aspekten der Weltwirtschaft ist es schwer zu erraten, welcher zuerst in einer Katastrophe enden wird. Zerstörung von 20 % der globalen Energieversorgung oder Weigerung, eine Superwaffe mit beispielloser Macht zu regulieren, um das Finanzsystem zu zerstören. Wer beißt zuerst? Oder werden sie beide?
Es gibt wahrscheinlich einige sehr selbstgefällige Menschen in Washington, die selbstgefällig sind, dass dies Amerikas Superwaffe ist, und nicht wissen, was Anthropic heute hat, was China und andere bald haben werden.
Die USA rufen Bankchefs wegen Cyberrisiken durch das neueste KI-Modell von Anthropic vor
The U.S. government warns financial institutions that Anthropic’s “Mythos” AI can find and exploit software vulnerabilities at an unprecedented scale, outperforming top humans and posing systemic risks to banks and the broader financial system.
byu/lughnasadh inFuturology
10 Kommentare
Am I supposed to trust anything the US government says at this point?
(Also it’s unfortunate that the warning about rule number one doesn’t just tell me what that rule is 😆)
Priming for the false financial collapse they’re trying to bring on the world.
But they’re not currently releasing mythos to the public so isn’t this a good thing? Let mythos find all the exploits in the systems and ensure the banks plug those holes.
That only means that government and criminals already have it.
Gatekeeping it from the public must have other motives.
According to your first reference, Claude is working with over 40 organizations to fix those critical bugs before Mythos is released. So despite complacent people in Washington, there are people who recognize the seriousness of this and are working to fix it.
The real story here is how overconfident humans are about our intelligence.
The spooky security armageddon is a marketing push. The capabilities you posited just don’t exist.
We should be realistic about the actual difficult security problems in our system, rather than doing a bunch of heavy lifting for a company that is currently in a panic about attempting to go public.
So the new model is identifying and exposing existing vulnerabilities that could already be getting exploited, regardless of any „zero day“ label. Sounds great! Thanks for the good news. Maybe developers will be able to take advantage of this functionality to avoid releasing even more massive exploits that last in the wild for years.
Idk, all I can say is that yesterday:
1. Neither I nor my friend could pay our metro phone bill online, as the system insisted that neither of our phone numbers is a legitimate metro phone number;
2. Two separate ATMs refused to dispense any cash to him without explanation, despite him having plenty of money in his account;
3. My roommate tried to buy groceries online at Safeway, and *every single item* in her cart was „unavailable.“
Don’t know what was going on, but it was definitely something. I’m certain it’s not related, but it did kind of give me a small indication of the kind of havoc these things can bring about.
Here’s the burning question I have. How long until the open source community catches up? What risk will this pose if something is not done fast enough to patch every single vulnerability, and can that race to patch win the open source race? Is this just an inevitability that our secure systems are going to become obsolete? Will we have to go back to paper again if this happens? If so maybe we should stop developement immediately due to the ability of it making us regress backward technologically 50 years.