BREAKING
Technology

OpenAI’s GPT-6 Cyber reportedly bypasses Apple’s Lockdown Mode

OpenAI’s latest cybersecurity model is drawing attention after reportedly identifying dozens of security weaknesses across Apple’s product ecosystem, including a way to get around some protections offered by the company’s Lockdown Mode.

According to a report by Forbes, GPT-6 Cyber has helped researchers uncover around 35 vulnerabilities affecting iPhones, Macs, iPads, Apple Watches, Apple TVs and Vision Pro headsets. Some of these flaws could potentially allow attackers to gain control of a device by simply getting a user to visit a specially designed malicious website.

OpenAI introduced GPT-6 Cyber on November 5, positioning it as an advanced tool for identifying and testing security vulnerabilities. The model received a ‘critical’ cybersecurity rating, reflecting its ability to discover flaws and develop potential exploits.

The research was reportedly carried out over several months by OpenAI and cybersecurity firm Calif. One of the more concerning discoveries involved Apple’s Lockdown Mode, a security feature designed to protect people who could be targets of highly sophisticated cyberattacks.

Researchers reportedly used GPT-6 Cyber to develop an exploit that can bypass some of the restrictions, in Lockdown Mode. Lockdown Mode is designed to limit device functions and reduce possible entry points for attackers. The discovery highlights that specialised security protections can have weaknesses.

Apple has reportedly fixed several flaws

The findings extend across Apple’s hardware and software, although the company has not publicly confirmed all 35 vulnerabilities.

Calif CEO Thai Duong told Forbes that Apple had fixed many of the reported security issues after receiving details from the researchers. Apple has not publicly responded to Forbes’ request for comment.

There is, however, some confirmation of OpenAI’s security research in Apple’s own documentation. In its October 26 security notes for iOS 27.2 and iPadOS 27.2, Apple credited OpenAI Codex Security for three WebKit vulnerabilities. It also acknowledged OpenAI alongside STAR Labs for a kernel vulnerability.

Two of the WebKit flaws could potentially allow arbitrary code execution when a device processes malicious web content. The same four credits also appear in Apple’s macOS Golden Gate 27.2 security notes.

However, these public acknowledgements do not confirm all 35 vulnerabilities mentioned in the report. Apple specifically credited OpenAI Codex Security, which should not be confused with GPT-6 Cyber.

What makes GPT-6 Cyber important?

GPT-6 Cyber is designed to help cybersecurity researchers identify weaknesses, test possible attacks and develop solutions before hackers can exploit them.

Its reported work is not limited to Apple products. OpenAI says the model helped identify 81 previously unknown high-severity vulnerabilities in Firefox, prompting Mozilla to begin addressing them.

The company also says it helped develop working exploits for 948 Linux kernel vulnerabilities, including 588 previously unknown flaws. Additionally, OpenAI shared 304 critical findings with Linux maintainers, along with proposed patches.

For now, GPT-6 Cyber remains restricted to approved customers through OpenAI’s Daybreak Red programme rather than being available to the general public.