It seems Apple has decided to recruit a new "genius" to its security team, but this time it's not a human. Anthropic recently announced a new initiative called "Project Glasswing," aimed at helping tech giants discover and fix security vulnerabilities before malicious actors do. The idea is both simple and frightening: using a new artificial intelligence model called Mythos Preview that outperforms even the most skilled human programmers in identifying software weaknesses.
Project Glasswing: The bug bounty that never sleeps
Project Glasswing is not just a technological experiment; it's an urgent attempt to put AI capabilities to the defense side of defense, rather than offense. According to Anthropic, the Mythos Preview model has already successfully discovered thousands of zero-day vulnerabilities—vulnerabilities that companies are unaware of—that existed in all the major operating systems and web browsers we use daily.
Artificial intelligence has now reached a level of sophistication in writing and reading code that surpasses human capabilities. With the accelerating pace of development, it has become essential for major companies like Apple to possess these defensive tools before blackmailers and hackers, who disregard security and privacy standards, acquire them. It's as if we're in an arms race, but the weapons here are lines of code and programming logic.
Apple and the Grand Security Alliance
Apple isn't alone in this exclusive club; access to the Mythos Preview won't be publicly available (lest our vulnerabilities end up being sold on the market), but rather limited to a select group of partners. This list includes, in addition to Apple, companies like Google, Microsoft, Amazon, and Nvidia. Anthropic is being very generous here, allocating up to $100 million in usage credits to its partners. After that, Apple will have to foot the bill to keep this "smart inspector" running.
Apple will use this technology to enhance the security of Safari and its various operating systems, from iPhones to macOS, watchOS, and even VisionOS. The goal is for the AI to read millions of lines of code in seconds, telling engineers in Cupertino: “Sorry, there’s a vulnerability here that could allow hackers in. It’s best to close it now before it’s too late.” Given the rumors surrounding a new homeOS, it’s a good idea for this system to begin its life under the watchful eye of robust AI security.
Source:
10 comment