A security firm reveals an exploit for Apple's M5 chip developed with AI

A cybersecurity startup has announced that it has created a working exploit for the macOS kernel, bypassing the protections of Apple's new M5 chip and the Memory Integrity Enforcement system. The discovery, made public through a post on Substack, represents a significant challenge to the security defenses of the Cupertino colossus, traditionally considered among the most robust in the industry.

Quick Response

  • The exploit targets the macOS 26 kernel on M5 hardware, exploiting two vulnerabilities and additional techniques
  • Development took less than a week, with the help of a preview version of Anthropic's AI Claude Mythos
  • The attack starts from a non-privileged user account and escalates to root privileges
  • Apple has not yet publicly commented on the claims
  • Apple's MIE protections use memory marking technology to prevent memory corruption attacks

The Vietnamese company Calif claims to have developed the first public macOS kernel memory corruption exploit capable of surviving Memory Integrity Enforcement (MIE) protections on M5 hardware. According to the company, the exploit was created in less than a week, with the help of a preview version of Claude Mythos, an AI model by Anthropic designed to identify and exploit software vulnerabilities.

Technical details of the exploit

The exploit targets macOS 26 on Apple M5 systems. According to Calif, the attack starts from a local non-privileged user account and escalates to root privileges using standard system calls. The exploit chain combines two vulnerabilities and additional techniques targeting the bare-metal M5 hardware with MIE enabled.

The company stated that the preview version of Mythos helped identify the vulnerabilities and assisted in the development of the exploit. However, Calif emphasized that human experience was still necessary to bypass Apple's new MIE protections.

The discovery is particularly significant because memory corruption bugs remain one of the most common methods used by attackers to compromise operating systems and applications. These bugs can allow attackers to crash a program, steal data, or even take control of a system.

The role of Claude Mythos in identifying vulnerabilities

Anthropic released the preview version of Mythos in April after internal and external tests suggested that the model could identify and exploit software vulnerabilities autonomously at a higher level than previous public AI models. Instead of releasing it publicly, Anthropic limited access to selected technology companies, banks, and researchers through its Project Glasswing initiative.

Mythos has demonstrated impressive capabilities in various scenarios. Mozilla reported that Mythos identified 271 vulnerabilities in Firefox during internal tests. The UK's AI Security Institute found that the model could autonomously complete sophisticated, multi-stage cyberattack simulations.

Implications for Apple's security

Calif described the M5 exploit as "a taste of what's coming." The company suggested that Apple's MIE protections were developed in an era before the existence of advanced AI models like Mythos Preview. This implies that current defenses may not be sufficient to address future AI-based threats.

The revelation raises important questions about the defensive capabilities of existing security technologies in the face of advancing AI offensive capabilities. As AI models become more sophisticated, they may be able to identify and exploit vulnerabilities at an unprecedented rate, making it harder for hardware and software manufacturers to keep up.

The future of AI-based exploits

The use of AI like Claude Mythos in developing exploits represents a significant shift in the cybersecurity landscape. While exploits were primarily the result of manual research conducted by human experts in the past, the integration of advanced AI models could significantly accelerate the process of discovering and exploiting vulnerabilities.

This development underscores the importance for companies to invest in advanced security technologies and adopt a proactive approach to vulnerability management. As attackers use more powerful tools, defenses must also evolve to maintain an advantage.

While the security community continues to monitor developments, Calif's revelation serves as a reminder that even the most secure systems can be vulnerable in the face of advancing technology. As we enter the era of AI-based exploits, the need for innovative and adaptable security solutions has never been more critical.

The broader context of cybersecurity

This exploit is not just a threat to Apple devices. It reflects a broader trend in the cybersecurity industry: the arms race between defenders and attackers is accelerating thanks to the adoption of advanced AI technologies. According to the Ponemon Institute, 68% of technology companies reported an increase in memory corruption vulnerability-based attacks over the past year. The integration of AI tools like Mythos Preview could significantly reduce the time required to identify and exploit these vulnerabilities, from months to weeks or even days.

Industry responses and possible countermeasures

The security community is already beginning to respond. Microsoft has announced a new security framework called "AI Shielded Execution," designed to protect critical processes from AI-based exploits. Similarly, Google has introduced "Project Zero for AI," an initiative that uses machine learning models to identify potential vulnerabilities before they can be exploited. However, these tools are still in the early stages of development and may take time to become effective against sophisticated threats like those identified by Calif.

Implications for end users

For Apple device users, this discovery raises critical questions about the security of their systems. While Apple has not yet publicly commented on Calif's claims, users may want to consider additional preventive measures. These include regularly updating operating systems, using advanced antivirus software, and limiting user accounts with administrative privileges. It is also important to note that, despite the exploit, Apple devices remain among the most secure on the market thanks to their integrated hardware and software architecture.

The future of cybersecurity in the age of AI

The emergence of AI-based exploits like the one developed by Calif suggests that the future of cybersecurity may be characterized by a greater reliance on advanced technologies. As companies continue to invest in research and development, we are likely to see an increase in AI-based solutions designed to identify and mitigate threats in real-time. However, this approach also presents challenges, including the potential risks of false positives and the need for human oversight to ensure the accuracy and effectiveness of AI solutions.

Ethical and regulatory considerations

Calif's revelation also raises important ethical and regulatory questions. As the capabilities of advanced AI tools advance, it becomes increasingly difficult to draw the line between legitimate research and potential abuse. Regulatory authorities may need to consider new rules to govern the use of these technologies, balancing the need for innovation with the protection of user privacy and security. Additionally, the international community may need to collaborate to develop shared standards for cybersecurity in the age of AI.

The exploit developed by Calif represents a significant turning point in the cybersecurity landscape. While it highlights the vulnerabilities of current systems, it also serves as a reminder of the need for innovative and adaptable solutions to address future threats. As we enter the era of AI-based exploits, collaboration between researchers, companies, and regulatory authorities will be crucial to ensure that cybersecurity can keep pace with technological evolution.

Editorial Note and Disclaimer

The guides and content published on GoYou are the result of independent research and analysis activities, for informational, educational, and in-depth purposes.

GoYou does not constitute a journalistic publication or an editorial product pursuant to Law No. 62/2001 and does not engage in real-time information activities.

The GoYou project does not provide professional, technical, legal, or financial advice and disclaims all responsibility for the improper use of the information published.

In the Crypto sector, every investment involves risks: readers are invited to always inform themselves independently before making any decision.