Forescout Vedere Labs researchers used Claude to port a pre-authentication RCE exploit between WAGO PLC models, moving CVE-2021-31886 from a 750-852 to a 750-831 in a lab setting. It worked: arbitrary ARM code execution without credentials.
But the friction tells the real story. Eight hours 32 minutes of continuous researcher guidance. $535.74 in API costs. The breakthrough wasn’t Claude alone: researchers had to identify that sending USER+CWD without CRLF termination would keep the shellcode in memory long enough to execute. Claude then iterated on that insight.
An attempt to escalate into a command-and-control implant permanently bricked the hardware. CERT@VDE confirmed no patches exist for affected WAGO devices. This isn’t proof that frontier models automate industrial exploitation unsupervised. It shows they can help research teams move faster on known, unpatched vulnerabilities that manufacturers won’t fix.
Sources: Forescout Vedere Labs
Coverage: The Hacker News
Related on this blog: [AUTO] AI-Assisted SharePoint Exploit Chain Reaches Unauthenticated RCE • [AUTO] UAT-10147 Uses AI to Automate Exploitation at Scale • [AUTO] AI-generated industrial exploits are now in the wild