In tech jargon they are called Bug Hunter. They are programmers, experts in computer sciencechild prodigy or specialized companies. Their job in recent years has been to sift through the software code and find bugsthose system flaws that can block a procedure or become the gateway for an expert hacker. Bug Hunters often work with companies that develop software. They find a flaw, get paid well, and before everything has been made public, the flaw is fixed. Or at least. It worked like that until he arrived Mythosthe latest artificial intelligence model developed by Claude.
Mythos is a model that has been talked about for a few days. When it was presented by Claude the announcement that accompanied the launch was not the most serene: the company leads from Dario Amodei has explained that it is not possible to open Mythos to just anyone. For the moment it remains confined to a few partners: according to the company its capabilities analyze bugs in a software and exploit them for a hacker attack are too sophisticated. In a post published on Anthropic’s site, the company explained that it has begun working with a long list of Big Tech companies to find bugs within their software with Mythos.
The project is called Project Glasswing and there are practically all the Big Techs: from Apple to Google, through Microsoft and NVIDIA. Just the ones are missing Elon Musk’s companies. On the other hand, Mythos, officially Mythos Previewwould have by now surpassed “humans most skilled at identifying and exploiting software vulnerabilities”. And again: “Myhtos has already identified thousands of highly severe vulnerabilities, some of which are present in all major operating systems and web browsers”.
Anthropic relies on religion: met 15 Christian leaders to teach AI to be more “good”
The Bugmageddon theory: when all the knots come to a head
Now, these bugs they are not found only in recent codes. Sometimes they drag on from years of updates. They don’t arouse suspicion, they don’t stand out but they are known. The fact that Mythos exists but has not been released to the public can be read as good news: The bugs he found will come resolved in environments protected, without giving anyone the opportunity to exploit them. However, there is a next step. Is there any other artificial intelligence that is capable of completing the same operation?
The Wall Street Journal coined the term Bugmageddonthe bug apocalypse. A phase in which, thanks to all the artificial intelligence models available, cyber criminals will be able to find, identify and exploit all remaining bugs in the software codes. If Mythos algorithms can do it now, other forms of artificial intelligence will also be able to do it in the future.
The Wall Street Journal reached out Niels Provosa computer scientist who 27 years ago had worked on the code of OpenBSD. Mythos algorithms have been working on this software in recent days, finding a bug that dates back to when Niels Provos had put his hands on the code: “To tell the truth, I found it simply hilarious. Because it’s very old code. Who knows when was the last time a human being even looked at it.” The consequence is clear, explains Provos: “Now the necessary level of competence to develop truly sophisticated exploits is drastically reduced”.
