On November 5, OpenAI rolled out its latest cybersecurity model, GPT-6 Cyber. The company touted the model as able to find and exploit security flaws, rating it ‘critical’ in cybersecurity.
Now the model has reportedly uncovered around 35 security vulnerabilities across Apple's products, including iPhones, Macs, iPads and even Apple Watches. Some of these flaws could potentially allow hackers to take control of devices through a malicious website.According to a report by Forbes, the findings came from research conducted by OpenAI and cybersecurity firm Calif over the past few months.
The researchers reportedly found vulnerabilities across Apple's hardware and software, with some serious enough to put users' devices at risk. They also used GPT-6 Cyber to develop an exploit that could bypass the protections offered by Apple's Lockdown Mode, a security setting designed for people who may be targeted by sophisticated cyberattacks.GPT-6 Cyber finds vulnerabilities across Apple devicesThe reported findings go beyond Apple's iPhones and Macs.
The report reveals that the researchers also discovered security weaknesses in Apple TVs and Vision Pro headsets.Some of the vulnerabilities could allow attackers to take control of an iPhone or Mac by getting a user to visit a malicious website. In simple terms, a seemingly harmless website could become a way for hackers to exploit a weakness in the device's software.
The Lockdown Mode finding is particularly interesting. Apple introduced this feature to protect users who may face highly targeted digital attacks.
When enabled, it restricts certain features and limits the ways attackers could access a device. However, researchers reportedly managed to use GPT-6 Cyber to write an exploit that could remove some of these protections.Calif CEO Thai Duong told Forbes that Apple fixed many of the vulnerabilities after the company was alerted.
However, Apple has not publicly confirmed the full list of 35 findings or responded to Forbes' request for comment. There is some public evidence of OpenAI's contribution, though.
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 four credits also appear in Apple's macOS Golden Gate 27.2 security notes.It is worth noting that Apple's public security credits do not provide a complete list of the 35 vulnerabilities mentioned in the Forbes report. They also specifically name OpenAI Codex Security, rather than GPT-6 Cyber, so the two should not be treated as interchangeable.What is OpenAI's GPT-6 CyberFor context, OpenAI introduced the GPT-6 Cyber model designed specifically for cybersecurity work.
It was touted as helping researchers identify security weaknesses, test how attackers could exploit them, and develop fixes before malicious hackers get a chance to exploit them.The model has already been put to work beyond Apple products. OpenAI says its researchers used it to identify 81 previously unknown high-severity vulnerabilities in Firefox, which Mozilla has since started fixing.
In another exercise, it helped develop working exploits for 948 Linux kernel vulnerabilities, including 588 that were previously unknown. OpenAI also says it shared 304 critical findings with Linux maintainers, along with proposed patches.For now, GPT-6 Cyber is not available to everyone.
OpenAI is restricting access to approved customers under its Daybreak Red programme.- Ends

