www.infosecurity-magazine.com 7/21/2026, 2:11:05 PM · external

Russian hacker sells AI tool bypassing Claude safety filters

Russian hacker sells AI tool bypassing Claude safety filters
CyberSIXT Evidence Panel Source marked as original reporting
Threat Actor
Trim

A Russian-speaking cybercriminal, known as 'Trim', transitioned from sharing a jailbreak tutorial on a Russian forum to selling a commercial AI pentest platform called 'AI Pentest Checker' within three months. This platform utilizes techniques outlined in his initial posts, including methods for bypassing Claude Opus' safety filters, such as 'Context Warming' and 'Ghost Reset'.

Trim sourced a grey-market Claude API key for $4 and integrated it with conventional scanning tools, claiming to conduct scans and generate reports in under ten minutes. The tool is characterized by its use of a leaked prompt that manipulates Claude's system, allowing for greater exploitation potential.

View full article

Article by CyberSIXT