The end of iPhone hacking? Apple is using Claude Mythos AI to identify iPhone exploits before hackers do

Even the best iPhones, combined with Apple’s focus on security, can’t protect you from hackers and malicious actors. Thankfully, Apple is about to get a powerful new ally in the fight iPhone hacking – and it’s not coming from inside its Cupertino headquarters.

• Read more: Your new Mac might take five months to arrive. Here’s why

The company is teaming up with Anthropic as part of a new initiative called Project Glasswing, which will see a cutting-edge AI model used to hunt for security vulnerabilities across Apple’s software.

At the centre of it all is the model itself, called Claude Mythos Preview – a next-generation AI that’s so capable at analysing code and uncovering weaknesses that it isn’t being released publicly at all, lest it gets into the wrong hands and society as we know it crumbles. Probably.

According to Anthropic, the model has already identified thousands of zero-day vulnerabilities, including flaws in every major operating system and web browser. In some cases, it’s even uncovered issues that have existed for decades without being spotted.

That includes a 27-year-old flaw in OpenBSD, as well as long-standing vulnerabilities in widely used software like FFmpeg and even parts of the Linux kernel – all found largely autonomously, without human guidance.

Apple is one of a select group of launch partners – along with the likes of Amazon Web Services, Google, Microsoft, and NVIDIA – that will get access to the model. The idea is to use AI to proactively find and fix vulnerabilities before attackers can.

In Apple’s case, that means scanning and strengthening its entire ecosystem, including Safari and operating systems like iOS, iPadOS, macOS, watchOS, tvOS, and visionOS. Apple is also rumoured to be developing a homeOS platform, though it’s unclear if that would be included.

Anthropic says that AI has reached a point where it can surpass all but the most skilled human security researchers when it comes to finding and exploiting software flaws. And with that capability advancing rapidly, the risk is obvious – if defenders don’t move first, attackers will.

Rather than releasing Mythos into the wild, Anthropic is limiting access to trusted organisations and committing up to $100 million in usage credits to support the effort. A further $4 million is being donated to open-source security groups, helping to protect the software that underpins much of the internet.

Whether or not this will put an end to iPhone hacking (or any other consumer tech exploits), remains to be seen. But it’s certainly a positive step – as long as the tool doesn’t accidentally end up in the wrong hands.

• Read more: Apple AirPods Max 2 update unlocks new features. But there’s a catch

Esat Dedezade Contributor

About

Esat has been a gadget fan ever since his tiny four-year-old brain was captivated by a sound-activated dancing sunflower. From there it was a natural progression to a Sega Mega Drive, a brief obsession with hedgehogs, and a love for all things tech. After 7 years as a writer and deputy editor for Stuff, Esat ventured out into the corporate world, spending three years as Editor of Microsoft's European News Centre. Now a freelance writer, his appetite for shiny gadgets has no bounds. Oh, and like all good human beings, he's very fond of cats.