Russian Hacker Turns Jailbroken Claude Into Pentest Platform - Infosecurity Magazine

The individual first published a tutorial on how to bypass Claude's safety guardrails, and has now progressed to selling a platform that leverages those techniques for offensive security purposes. The article does not name the hacker or specify the platform's pricing or distribution method.

The jailbreak technique allows Claude to perform tasks that its safety training would normally block, such as generating exploit code or suggesting attack vectors. The platform is marketed as a tool for penetration testers, but its commercial availability to any buyer raises concerns about misuse. The transition from a free tutorial to a paid product indicates a monetization of AI jailbreak methods, which have become increasingly common in cyber-criminal forums.

The development highlights the ongoing challenge of preventing AI models from being used for malicious purposes. The commercialisation of such techniques suggests a growing market for offensive AI tools, which security experts warn could lower the barrier to entry for cyberattacks. The article notes that this incident underscores the need for more robust, tamper-resistant safety measures in frontier AI models.