Forescout AI-Assisted Research Finds 66 New Flaws Across 10 TCP/IP Stacks
Vulnerability Disclosure · 5 sources · Updated 1d ago · First seen Nov 16, 2026

Forescout researchers confirmed 99 vulnerabilities across 10 TCP/IP stacks, including 66 previously unknown flaws, using an AI-assisted research system combining Claude models, security tools and researcher-developed scanning rules. Of the newly discovered vulnerabilities, 52% were rated high or critical severity. The affected software powers industrial equipment, medical devices, networking products and other connected systems; Siemens identified affected APOGEE, Desigo and TALON building automation controllers. Following system development and environment preparation, the research required approximately 10 hours of automated analysis and 80 hours of manual validation, compared with roughly nine months for the earlier research effort.
Estimated model costs totaled approximately $1,200–$2,300, but compiled-software analysis required substantially more resources than source-code analysis, and the system missed many known vulnerabilities. Several projects have issued fixes, while investigations into three networking stacks remain undisclosed. Discontinued devices without vendor patch support present a persistent containment challenge, particularly in critical infrastructure. Forescout recommended risk-based remediation, reduced device exposure, isolation of unpatchable systems, faster patching and stronger supplier security requirements as AI accelerates vulnerability discovery for both defenders and attackers.


