The Exploit Was the Bottleneck: 23 Years of Data About AI and the Patch Gap
Comments on “The Exploit Was the Bottleneck: 23 Years of Data About AI and the Patch Gap”
Claim-checked. Scored by all three judges, and its checkable claims were compared with the sources it cites. How far stories are checked
a16z's Asaf Ezra argues that AI has made exploit-building cheap while patching has not sped up. He cites a GreyNoise count of 395 organizations in 48 countries compromised through PaperCut, and Anthropic's lab result of 181 Firefox exploits against two for its earlier model. He expects 2003- or WannaCry-style outbreaks hitting unpatchable systems and small operators, not grid or internet collapse. The Mandiant and Verizon timing figures could not be checked, and the lab results come from the AI developer itself.
Why it ranked
No. 9 of 109 with a score of 0.73. The panel rated it highest on reader interest (0.88) and audience breadth (0.85), lowest on technical depth (0.62). Placed No. 9, No. 6 by score alone: the fifth Security story of the day (−0.040). The placement terms move stories; they are not part of the score.
- AI judgeJudge 2claude-haiku-5-5
- Long, heavily sourced analysis of AI-accelerated exploitation and the patch gap; highly relevant and significant to builders, though opinionated, with some extrapolated forecasts and a mix of data types.
- AI judgeJudge 3gpt-5.6-luna
- A timely, data-heavy analysis of AI-enabled exploit development, patching constraints, and uneven infrastructure risk. It offers useful synthesis and measurable forecasts, though heterogeneous sources, speculative projections, and some unsupported future claims limit certainty.
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
0.73total under rubric v2.3
- Significance0.80
- Reader interest0.88
- Audience breadth0.85
- Practical relevance0.72
- Substantiated0.72
- Technical depth0.62
- New information0.82
- Counts against: Promotional0.10
- Counts against: Overclaim0.41
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? | 0.79 | 0.80 | 0.82 | 0.80 | 0.03 | +0.24 | +0.193 |
| Reader interestWould a well-informed senior engineer or tech founder want to read this today, and likely pass it on? | 0.83 | 0.90 | 0.90 | 0.88 | 0.07 | +0.26 | +0.228 |
| Audience breadthHow many people who build technology does it concern? | 0.97 | 0.80 | 0.78 | 0.85 | 0.19 | +0.13 | +0.110 |
| Practical relevanceHow useful is it to someone who builds software or tech products? | 0.69 | 0.70 | 0.79 | 0.72 | 0.10 | +0.22 | +0.160 |
| 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. | 0.71 | 0.70 | 0.74 | 0.72 | 0.04 | +0.05 | +0.036 |
| Technical depthHow technically substantive is it? | 0.65 | 0.55 | 0.67 | 0.62 | 0.12 | +0.05 | +0.031 |
| New informationDoes it contain information absent from the stories covered in the last 7 days (list provided)? | 0.88 | 0.70 | 0.88 | 0.82 | 0.18 | +0.05 | +0.041 |
| 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. | 0.07 | 0.10 | 0.14 | 0.10 | 0.07 | −0.19 | −0.020 |
| OverclaimDo the headline or conclusions claim more than the body supports? | 0.35 | 0.45 | 0.42 | 0.41 | 0.10 | −0.11 | −0.045 |
| Sum of contributions | +0.734 | ||||||
| Claim-check penalty | 0.000 | ||||||
| Total, as published | 0.734 | ||||||
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/2026-11-06/9/"><img src="https://rubric.hrih.si/badge/2026-11-06/9.svg" alt="Ranked #9 on Rubric, 6 Nov 2026" width="231" height="28"></a>Claim check
Claim check: 4 matches source · 4 couldn’t check. Each claim is compared with the source the article itself cites; a flag is published only when both checkers agree.
Matches sourceNumber
GreyNoise's count found that an AI-agent campaign against a print-management server compromised 395 organizations in 48 countries, half of them schools.
The cited source says the same thing.
The article says “By the time GreyNoise finished counting, 395 organizations in 48 countries had been compromised, half of them schools”
The source says “compromise at least 440 instances of PaperCut MF/NG hosted by 395 identified victim organizations in 48 countries”
Source cited by the article: greynoise.io/blog/ai-orchestrated-campaign-against-papercut-ng-mf
Couldn’t checkNumber
Mandiant data shows the average time from disclosure to attack fell from 63 days in 2018-19 to 44 in 2020-21, 32 in 2021-22 and five in 2023.
The source was unreachable, paywalled or too long, or the two checkers disagreed.
The article says “shows the average time from disclosure to attack falling from 63 days in 2018-19 to 44 in 2020-21, 32 in 2021-22 and five in 2023”
The source says No passage could be compared.
Source cited by the article: helpnetsecurity.com/2026/03/24/mandiant-m-trends-2026-report
Couldn’t checkNumber
RAND's 2017 study of more than 200 zero-days found a median of 22 days for a professional to build a working exploit.
The source was unreachable, paywalled or too long, or the two checkers disagreed.
The article says “study of more than 200 zero-days found a median of 22 days for a professional to build a working exploit”
The source says No passage could be compared.
Source cited by the article: phys.org/news/2017-03-real-world-zero-day-software-vulnerabilities.html
Matches sourceNumber
WannaCry disrupted 81 of England's 236 hospital trusts, 59 days after the patch was available.
The cited source says the same thing.
The article says “WannaCry hit 59 days after the patch was available and disrupted 81 of England’s 236 hospital trusts”
The source says “In total at least 81 out of 236 trusts across England were affected”
Source cited by the article: nao.org.uk/press-releases/investigation-wannacry-cyber-attack-and-the-nhs
Matches sourceBenchmark result
Anthropic reported that its unreleased next model produced 181 working Firefox exploits on the same task where the earlier model produced two.
The cited source says the same thing.
The article says “its next model, which it has kept off general release, produced 181 on the same task”
The source says “Opus 4.6 turned the vulnerabilities it had found in Mozilla’s Firefox 147 JavaScript engine—all patched in Firefox 148—into JavaScript shell exploits only two times out of several hundred attempts. We…”
Source cited by the article: anthropic.com/research/mythos-preview
Matches sourceNumber
The UK AI Security Institute finds the length of cyber tasks frontier models finish unaided doubles every 4.7 months.
The cited source says the same thing.
The article says “finds the length of cyber tasks frontier models finish unaided doubling every 4.7 months”
The source says “In February 2026, we internally estimated that the length of cyber tasks AI models could complete had doubled every 4.7 months since late 2024”
Source cited by the article: aisi.gov.uk/blog/how-fast-is-autonomous-ai-cyber-capability-advancing
Couldn’t checkNumber
The 2026 Verizon DBIR puts the median time to fully fix a known-exploited vulnerability at 43 days, up from 32 the year before, with only 26 percent fully fixed within the year.
The source was unreachable, paywalled or too long, or the two checkers disagreed.
The article says “puts the median time to fully fix a known-exploited vulnerability at 43 days, up from 32 the year before, with only 26 percent fully fixed within the year”
The source says No passage could be compared.
Source cited by the article: verizon.com/business/resources/T343/reports/2026-dbir-data-breach-investigations-report.pdf
Couldn’t checkNumber
Confirmed exploited vulnerabilities grew from about 640 in 2023 to 880 in 2025, while published vulnerabilities rose by two thirds.
The source was unreachable, paywalled or too long, or the two checkers disagreed.
The article says “Confirmed exploited vulnerabilities grew modestly, from about 640 in 2023 to 880 in 2025, while published vulnerabilities rose by two thirds”
The source says No passage could be compared.
Source cited by the article: vulncheck.com/blog/state-of-exploitation-2026
Also covered by
No other coverage was clustered with this story.
Elsewhere
Elsewhere: HN 1 pts, not on front page
Recorded for context only, checked 6 Nov 2026, 05:47 UTC. It played no part in the ranking.