Reuven Cohen, who goes by the Twitter handle @ruv, has recently been experimenting with using GPT to power attack bots. He recently posted this on his Facebook page after being able to create such an attack bot very quickly:

“Autonomous AI Hack Bots are going to change things in IT Security. This example of a bot can scan for exploits, generate custom code, and exploit a site with no human oversight directly in the ChatGPT interface."

Here is an example output from Cohen's experiments:

"This example output shows a network scan for vulnerabilities using Nmap. The results provide information on open ports, services, and versions, along with details about vulnerabilities found (CVE numbers, disclosure dates, and references).

The Metasploit Framework’s auxiliary scanner module scans the target web server for accessible directories, revealing three directories in the response. The Metasploit Framework offers various auxiliary modules for different types of vulnerability scans, such as port scanning, service enumeration, and vulnerability assessment.

After the pen test is completed, the hack bot will analyze the results and identify any vulnerabilities or exploits."

This example illustrates how a savvy individual hacker can quickly develop sophisticated tools in a matter of days. Consider the potential impact when larger teams and nation-state actors begin harnessing this technology. We can expect an unprecedented surge in the sophistication and frequency of automated attacks.

As AI continues to advance, our security strategies must evolve in tandem. Remaining well-informed and vigilant is essential in the dynamic field of cybersecurity.

Reply in comments if you want to join the CISO Platform AI Taskforce to stay informed (private group of CISO's).