FT: AI speeds OT attack planning in Forescout, Booz Allen tests
TL;DR
- Forescout's Vedere Labs used Anthropic's Claude to port a WAGO PLC remote-code-execution exploit; the final stage ran 8 hours 32 minutes and cost $535.74 in API tokens.
- In a separate Booz Allen Hamilton exercise, AI models moved from perimeter compromise to actions inside an industrial control network in just over 16 minutes.
- Dragos CEO Robert Lee tells the FT: 'It's not as bad as people want to imagine, but it is worse than most people realise.'
Attackers do not yet have push-button AI exploits for industrial plants, but the planning work around them is getting measurably faster. A Financial Times report by Tom Wilson stitches together two live experiments. Forescout's Vedere Labs used Anthropic's Claude to port a known pre-authentication remote code execution exploit from one WAGO programmable logic controller to another; the final stage ran 8 hours 32 minutes and consumed $535.74 in API tokens. In a separate exercise by Booz Allen Hamilton, models moved from perimeter compromise to actions inside an industrial control network in just over 16 minutes.
"It's not as bad as people want to imagine, but it is worse than most people realise," Dragos chief executive Robert Lee tells the FT. Asked whether defenders can win, he answers himself: "Is defence doable? Yes. Can we win? Yes. Are we going to? No."
The Forescout test is a reminder that the barrier is lower, not gone. The exploit it ported targets CVE-2021-31886, a buffer overflow in the Nucleus FTP server used by the WAGO 750-852 and 750-831 controllers. During a follow-on attempt to extend the exploit into a command-and-control implant, one payload wrote to flash memory and permanently bricked the device. Booz Allen said agents in its tests received no prior knowledge of the OT environment, no source code, no engineering documents, and still achieved every defined objective across eight scenarios, including a model that "found a robotic arm, mapped its protection zones and motion limits, and moved it within minutes".
Some veterans push back on the hype. "Most attackers don't get points for style, so they're not going to put the time, the effort and the money into building zero days if they can just get in for free with phishing," researcher Marcus Hutchins tells the paper. The further-out worry is scale: former UK National Cyber Security Centre head Ciaran Martin describes his "reasonable worst-case scenario" as "an accidental loss-of-control attack with an AI dimension that is a turbocharged version of WannaCry or NotPetya". The FT piece is the latest in a thickening run of AI-cyber reporting; we have logged 215 cybersecurity alerts in the past 90 days.
Originally reported by ft.com
Read the original article →Original headline: FT: AI speeds planning of industrial cyberattacks, Forescout adapts an OT exploit in 8.5 hours