Finding Critical Infrastructure Bugs with Off-the-Shelf AI
Comments on “Finding Critical Infrastructure Bugs with Off-the-Shelf AI”
Claim-checked. Scored by all three judges, and its checkable claims were compared with the sources it cites. How far stories are checked
Forescout built an AI harness for embedded software, combining Joern static rules, Claude models and emulation-based validation, and ran it against 10 TCP/IP stacks. It reports 99 vulnerabilities in 90 hours, 66 of them new, at about $935 for static analysis. It rediscovered 40% of its earlier manual findings, versus nine months of manual work then. Binary targets cost six times more than source code. The harness is strongest on memory-safety bugs and weaker on logic flaws. The cost comparison excludes harness-building effort, and the figures come from Forescout alone.
Why it ranked
No. 3 of 123 with a score of 0.76. The panel rated it highest on new information (0.91) and technical depth (0.90), lowest on audience breadth (0.72). Placed No. 3, No. 2 by score alone: close to “AI agents stole 190k gaming credentials for under $400 in model costs”, No. 14 on Monday 9 November 2026 (−0.004). The placement terms move stories; they are not part of the score.
- AI judgeJudge 2claude-haiku-5-5
- Strong security research with concrete cost and time metrics on AI-driven vulnerability discovery in TCP/IP stacks; detailed methodology, though much is vendor-report style and some claims are only partly verified.}
- AI judgeJudge 3gpt-5.6-luna
- Primary-style security research reports a reproducible AI-assisted harness, 99 vulnerabilities, costs, timing, precision, methodology, and mitigation. It is highly actionable and technically deep; some comparisons and broader attacker implications remain inferential.
Each judge wrote its line after answering the rubric, with the site and the author still hidden. Every answer behind this is in the table under Score.
Score
0.76total under rubric v2.3
- Significance0.74
- Reader interest0.84
- Audience breadth0.72
- Practical relevance0.81
- Substantiated0.89
- Technical depth0.90
- New information0.91
- Counts against: Promotional0.06
- Counts against: Overclaim0.30
Each judge’s answer per question (0–1), their mean and spread, and what the question adds to the score (weight × mean).
| Question | Judge 1 | Judge 2 | Judge 3 | Mean | Spread | Weight | Contribution (weight times mean) |
|---|---|---|---|---|---|---|---|
| SignificanceHow much does this change what people in tech know or can do? | 0.75 | 0.60 | 0.86 | 0.74 | 0.26 | +0.24 | +0.177 |
| Reader interestWould a well-informed senior engineer or tech founder want to read this today, and likely pass it on? | 0.78 | 0.80 | 0.95 | 0.84 | 0.17 | +0.26 | +0.219 |
| Audience breadthHow many people who build technology does it concern? | 0.83 | 0.55 | 0.78 | 0.72 | 0.28 | +0.13 | +0.094 |
| Practical relevanceHow useful is it to someone who builds software or tech products? | 0.84 | 0.70 | 0.90 | 0.81 | 0.20 | +0.22 | +0.179 |
| SubstantiatedHow well is it backed? Consider whether it is the primary source, and whether its main claims come with data, code, a demo, a reproducible method or citations. | 0.95 | 0.80 | 0.92 | 0.89 | 0.15 | +0.05 | +0.045 |
| Technical depthHow technically substantive is it? | 0.97 | 0.75 | 0.96 | 0.90 | 0.23 | +0.05 | +0.045 |
| New informationDoes it contain information absent from the stories covered in the last 7 days (list provided)? | 0.90 | 0.85 | 0.99 | 0.91 | 0.14 | +0.05 | +0.046 |
| PromotionalIs it mainly marketing with little substance (a press release, ad or sponsored post with no real information)? A launch post that carries real news is not promotional. | 0.06 | 0.10 | 0.03 | 0.06 | 0.07 | −0.19 | −0.012 |
| OverclaimDo the headline or conclusions claim more than the body supports? | 0.31 | 0.30 | 0.30 | 0.30 | 0.01 | −0.11 | −0.033 |
| Sum of contributions | +0.758 | ||||||
| Claim-check penalty | 0.000 | ||||||
| Total, as published | 0.758 | ||||||
Spread is the highest minus the lowest judge answer. marks a spread above 0.40: the judges disagree, and the story is marked contested.
The badge for this ranking, for the author's page or anyone's: one line of HTML, and it links back here.
<a href="https://rubric.hrih.si/editions/2026-11-14/3/"><img src="https://rubric.hrih.si/badge/2026-11-14/3.svg" alt="Ranked #3 on Rubric, 14 Nov 2026" width="238" height="28"></a>Claim check
Claim check: 1 matches source · 6 no source given · 1 couldn’t check. Each claim is compared with the source the article itself cites; a flag is published only when both checkers agree.
No source givenNumber
Forescout's AI harness found 99 vulnerabilities across 10 TCP/IP stacks in 90 hours, of which 66 were new.
The article gives no link or citation for it.
The article says “This research uncovered 99 vulnerabilities in 90 hours including automated analysis and human validation”
The source says No source to compare with.
No source givenBenchmark result
The harness rediscovered 33 of the 82 Project Memoria+ vulnerabilities, or 40%.
The article gives no link or citation for it.
The article says “Out of the 82 vulnerabilities in Project Memoria+, the harness rediscovered 33 (40%”
The source says No source to compare with.
No source givenNumber
Total static analysis cost was about $935 for 362 candidate findings, an average of $2.58 per finding.
The article gives no link or citation for it.
The article says “the pipeline processed 362 candidate findings for roughly $935, an average of $2.58 per finding”
The source says No source to compare with.
No source givenCausal conclusion
AI cut analysis time from about 9 months to 90 hours, and the $1,200 to $2,300 model spend is likely 100x less than a senior researcher's manual analysis cost.
The article gives no link or citation for it.
The article says “AI cut the analysis time from 9 months to 90 hours. But the cost difference is even more striking: the $1,200 to $2,300 of model spend for this study is…”
The source says No source to compare with.
No source givenNumber
Static analysis of binary targets was on average six times more expensive and took twice as long as source-code analysis.
The article gives no link or citation for it.
The article says “Static analysis of binary stacks was on average six times more expensive and took twice as long”
The source says No source to compare with.
Matches sourceNumber
Project Memoria started in 2020, took 18 months and uncovered almost 100 CVEs in 14 embedded TCP/IP stacks.
The cited source says the same thing.
The article says “That project started in 2020, took 18 months and uncovered almost 100 CVEs in 14 stacks used by embedded devices of dozens of vendors”
The source says “Over the past 18 months we have announced 5 different phases of vulnerability research with a total of 97 vulnerabilities, affecting 14 TCP/IP stacks”
Source cited by the article: forescout.com/research-labs/project-memoria
Couldn’t checkFirst, only or fastest
Siemens was the first vendor to identify vulnerable devices: APOGEE, Desigo and TALON building automation controllers running Nucleus NET are susceptible to the newly discovered issues.
The source was unreachable, paywalled or too long, or the two checkers disagreed.
The article says “Siemens was the first to identify vulnerable devices: several building automation controllers in the APOGEE, Desigo and TALON product lines run Nucleus NET”
The source says No passage could be compared.
Source cited by the article: productcert.siemens.com/productcert/html/ssa-333035.html
No source givenNumber
Of the 54 findings reported to vendors and maintainers, 20 affect open-source stacks and 34 affect proprietary stacks, and only five have been fixed to date.
The article gives no link or citation for it.
The article says “20 affecting open-source TCP/IP stacks , and 34 affecting proprietary stacks (only five have been fixed to date”
The source says No source to compare with.
Also covered by
Elsewhere
Elsewhere: HN 1 pts, not on front page
Recorded for context only, checked 15 Nov 2026, 20:44 UTC. It played no part in the ranking.