SIAgentNews.com
← All news

OpenAI model finds 35 security flaws in Apple devices

Researchers used a specialized AI model to uncover vulnerabilities in iPhones, Macs, and even bypass some of Apple's high-level security features.

· First reported by News9Live on MSN

Blueprint drawing: picture for the story 'OpenAI model finds 35 security flaws in Apple devices'
AI-made blueprint drawing for this story. No real people or products are shown.
35security vulnerabilities identified

The short version

  • Researchers identified approximately 35 vulnerabilities across iPhones, Macs, iPads, Apple Watches, Apple TVs, and Vision Pro headsets.
  • The study used a specialized AI model called GPT-6 Cyber to find weaknesses and test if they could be exploited by hackers.
  • The team developed an exploit capable of removing some protections offered by Apple's Lockdown Mode, a feature for high-risk users.
  • Apple has reportedly fixed many of the reported weaknesses after they were disclosed by the research team.

Why it matters

This story highlights how rapidly AI is accelerating the discovery of software bugs. It shows that even the most secure 'walled gardens' can have programming mistakes that AI can help uncover.

Cracks in the walled garden

Apple has built a long-standing reputation for securing its devices through a tightly controlled software environment often called a "walled garden." This ecosystem relies on strict App Store rules and specific security features to keep unauthorized users and hackers out of personal data. However, a recent investigation suggests that artificial intelligence is finding new ways to probe these defenses.

Researchers from OpenAI and the cybersecurity firm Calif used a specialized AI model to scan Apple's software for potential entry points. The study identified approximately 35 security vulnerabilities across a wide range of products, including iPhones, Macs, iPads, Apple Watches, Apple TVs, and Vision Pro headsets. These flaws represent programming mistakes that could potentially be used by malicious actors to compromise user systems.

The scope of the research is significant because it covers nearly every major piece of hardware in the Apple lineup. While some flaws are minor, others were severe enough to allow for remote compromise. For example, some vulnerabilities could allow a specially prepared webpage to become an entry point for an attack under specific conditions.

How GPT-6 Cyber works

The research relied on a tool called GPT-6 Cyber, which was introduced on November 5 as a specialized AI model for cybersecurity research. Unlike general-purpose chatbots, this model is designed to help experts examine complex software code to identify hidden weaknesses. It serves as a sophisticated scanner that can find bugs and then test whether those bugs can actually be exploited in a real-world scenario.

By using AI to automate the discovery process, researchers can scan much larger amounts of code than would be possible for human engineers alone. The model helped the team pinpoint specific weaknesses that could enable a system takeover on a Mac or a remote compromise on an iPhone. This represents a shift in how security auditing is performed, moving toward a model where AI acts as a primary tool for finding flaws.

The speed at which these models can identify bugs is a growing concern for software developers. As AI becomes more capable of offensive work, it creates a race between those trying to secure systems and those looking for ways to break them. The researchers used this capability to systematically map out the security landscape of Apple's current software offerings.

Bypassing Lockdown Mode

One of the most notable findings from the study involved Apple's Lockdown Mode. This is a high-security feature intended for individuals who face sophisticated cyberattacks, such as journalists or activists. It works by restricting certain device features, websites, and attachments to reduce the surface area available for spyware to exploit.

The researchers reportedly used GPT-6 Cyber to develop an exploit that could remove some of these specific protections. While this does not mean the mode has failed against every possible attack, it does suggest that the defenses are not impenetrable. It highlights how even the most rigorous security measures can have specific gaps that AI can help identify and exploit.

Despite this discovery, the CEO of Calif, Thai Duong, continues to recommend that people facing serious security threats use the feature. The goal of the research was to disclose these weaknesses so they could be addressed. The team's work serves as a stress test for Apple's most advanced security protocols, showing where the current protections might fall short.

Response and implications

Following the disclosure of these findings, Apple took action to address the issues. Calif reported to Forbes that the company fixed many of the reported weaknesses after the researchers shared their data. OpenAI also confirmed that Apple credited the company for identifying the vulnerabilities found during the process.

There is currently no confirmed evidence that any of the 35 vulnerabilities were actually exploited in attacks against ordinary users. However, the existence of these flaws proves that even tightly controlled systems contain human errors. The research underscores the reality that software security is an ongoing battle against evolving methods of discovery.

The broader implication is how quickly AI is changing the cybersecurity landscape. As frontier models get better at offensive work, they make it easier for researchers to find serious bugs faster. This same capability could eventually make life easier for cybercriminals, creating a new challenge for companies like Apple to keep up with AI-driven threats.

Staying secure today

For the average user, the discovery of these flaws does not mean that Apple devices are suddenly unsafe. Most users can maintain a high level of security by following standard best practices. The immediate advice for iPhone and Mac users is to keep their devices updated with the latest versions of iOS, iPadOS, and macOS.

Security updates are the primary way companies like Apple patch the types of vulnerabilities found by researchers. By enabling automatic updates, users ensure that they receive these fixes as soon as they are released. Additionally, users should continue to avoid clicking on suspicious links or opening unexpected attachments from unknown sources.

For those who are at a higher risk of targeted attacks, Apple's Lockdown Mode remains a useful additional layer of defense. While the research showed some bypasses, the mode still provides a significant barrier against many types of spyware. The key takeaway is to stay vigilant and keep software current to mitigate the risks posed by new discovery tools.

What to watch

  • How other tech companies respond to AI-driven security auditing.
  • Whether Apple releases a public patch specifically addressing the Lockdown Mode bypass.
  • The development of more specialized AI models for offensive cybersecurity.

Try it

Check your device settings to ensure that automatic software updates are turned on for your iPhone, Mac, and iPad.

Sources and links

First reported by News9Live on MSN. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.

Liked it? Say why

Add a line for other readers.

No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.

More AI news