Forescout AI harness finds 66 TCP/IP stack bugs
An AI harness on Claude models found 66 new vulnerabilities in embedded TCP/IP stacks used in IoT and industrial devices, for an estimated $1,200-$2,300 in tokens.
What happened
Forescout set out to test whether generally available models on a limited budget could find bugs in software relevant to critical infrastructure. It built an AI harness for embedded software (reverse engineering, security analysis and dynamic emulation) using Claude models and ran it against 10 TCP/IP stacks, seven from its earlier Project Memoria research and three new ones, including Contiki (low-power IoT devices such as street lights), Ethernut, lwIP and wolfIP (real-time systems in safety-critical devices).
The harness rediscovered 33 previously known vulnerabilities and found 66 new ones. Per Cybersecurity Dive, about half were memory-corruption bugs, 27% integer problems, 16% input-validation failures and 8% logic or control-flow flaws; Forescout says the harness is good at memory-safety bugs in parsing code and found logic bugs mostly by accident. Forescout reported 54 of the new findings (20 in open-source stacks, 34 in proprietary ones). Forescout’s post says only five had been fixed at publication; Cybersecurity Dive reports 24 fixed and 29 under investigation, with the Ethernut developer not responding. The other 12 were not reported because Forescout believed they were absent from current versions. The source does not list CVE numbers.
Cost and significance
Forescout estimates it spent $1,200-$2,300 on tokens, which it puts at roughly one hundredth of a senior vulnerability hunter’s salary (the company’s comparison). It warns that the affected stacks could reach “hundreds of products” and that discontinued equipment is especially exposed: the harness found seven flaws, including hardcoded admin credentials, in a stack used by an unsupported wireless access point. See the AI CVE tracker.
Timeline
Every dated event in the wiki that involves Forescout AI harness finds 66 TCP/IP stack bugs, newest first.
Forescout built an AI harness on Claude models for embedded software and ran it on 10 TCP/IP stacks used in IoT and industrial devices (Contiki, Ethernut, lwIP, wolfIP and others). It rediscovered 33 known bugs and found 66 new ones, about half memory corruption, for an estimated $1,200-$2,300 in tokens (Forescout's figure). It reported 54; Cybersecurity Dive counts 24 fixed so far.
Connections
Forescout AI harness finds 66 TCP/IP stack bugs and everything one link away. Drag to move, scroll to zoom, click a node to open it.Open in the full graph →
What links here
Concept AI and CVEs: latent flaws, now exploitable
Company Forescout