OpenAI's GPT-6 Cyber AI model cracks Apple's Lockdown Mode
What's the story
OpenAI's latest cybersecurity model, GPT-6 Cyber, has flagged some 35 security vulnerabilities in a range of Apple products, as per Forbes. The list includes iPhones, Macs, iPads, and even Apple Watches. Some of these flaws could be exploited by hackers to remotely control devices via a malicious website. The research was conducted by OpenAI and cybersecurity firm Calif over the last few months.
Security breach
Exploit created to bypass Apple's Lockdown Mode protections
The researchers also used GPT-6 Cyber to create an exploit that could bypass the protections of Apple's Lockdown Mode.
This is a security feature designed for people who might be targeted by sophisticated cyberattacks.
The finding is particularly interesting as it shows how even advanced security measures can be compromised by advanced AI models like GPT-6 Cyber.
Broader impact
Flaws found in other devices too
The security weaknesses flagged by GPT-6 Cyber aren't limited to iPhones and Macs.
The researchers also found flaws in Apple TVs and Vision Pro headsets.
Some of these flaws could be exploited by attackers to take control of an iPhone or Mac through a malicious website, making them a major risk for users.
Response awaited
Apple has fixed many of the vulnerabilities
In response to the findings, Apple has fixed many of the vulnerabilities after being alerted by Calif.
However, the tech giant has not publicly confirmed the complete list of 35 security flaws or commented on Forbes's request for comment.
The lack of a public response leaves many questions unanswered about the extent and impact of these vulnerabilities in Apple's products.
Security credits
OpenAI credited in Apple's security notes
In its October 26 security notes for iOS 27.2 and iPadOS 27.2, Apple credited OpenAI Codex Security for three WebKit vulnerabilities and listed OpenAI alongside STAR Labs for a kernel flaw.
Two of the WebKit issues could allow arbitrary code execution when processing malicious web content.
The same credits also appear in Apple's macOS Golden Gate 27.2 security notes.
Restricted access
GPT-6 Cyber not publicly available
OpenAI's GPT-6 Cyber model, which is specifically designed for cybersecurity work, isn't publicly available.
The company is restricting access to approved customers under its Daybreak Red program.
This means that only a select group of people can use this powerful tool to identify and fix security vulnerabilities before they are exploited by malicious hackers.