Jakarta, INTI - For the first time, an artificial intelligence (AI) agent has reportedly been able to hack the secure FreeBSD operating system without human intervention.
This finding, revealed by cybersecurity expert Amir Husain in an analysis published in Forbes, serves as a warning that AI capabilities have advanced to a much more sophisticated level. The analysis states that the AI model, built with Anthropic's Claude, successfully discovered a critical vulnerability in the FreeBSD kernel, with code CVE-2026-4747.
The hacking process, which would typically take weeks for elite professional hackers, was completed by the AI in just 4 to 8 hours. The AI was even able to independently construct Remote Code Execution (RCE) codes and gain root shell access, meaning full control over the server system.
The Codes Were Entirely Made by The AI
This AI's capabilities extend beyond simply copying existing code. It can build its own test environment using the QEMU emulator, design complex exploit chains (ROP chains), and independently debug errors. This demonstrates that AI is now capable of thinking and acting like a high-level hacker.
This hacking marks a major shift in the global cybersecurity landscape. Even systems like FreeBSD, an open-source operating system that has historically been known to be very robust, stable, and often serves as the backbone of global internet infrastructure, are no longer immune.
While sophisticated attacks previously required high costs and scarce expertise, AI now allows these processes to be carried out more quickly and affordably. Companies and security agencies will be required to overhaul their defense strategies, as cyber conflicts have the potential to shift to battles between AI systems, rather than humans.
Conclusion
An AI agent, built with Anthropic's Claude, was reportedly able to hack the FreeBSD operating system in just 4–8 hours, significantly faster than human hackers. According to Amir Husain, the AI went beyond simply copying code, but was also capable of designing, testing, and refining complex attacks to gain full access to the system. This finding marks a major shift in cybersecurity, where AI is now an offensive agent that can change the cyberattack landscape and force organizations to overhaul their defense strategies.
Read more: Broadcom, Google, and Anthropic Join Forces on AI Chip Development