--- title: After 7 Months, Glasswing Overwelms Maintainers with... description: Seven months into Anthropic's Project Glasswing, VulnCheck finds the results don't match the launch hype. Only 442 CVEs are attributable to Anthropic, about... canonical: https://daily.dev/posts/after-7-months-glasswing-overwelms-maintainers-with-underwhelming-findings-hljfurpi4 twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: After 7 Months, Glasswing Overwelms Maintainers with Underwhelming Findings | daily.dev og:description: Seven months into Anthropic's Project Glasswing, VulnCheck finds the results don't match the launch hype. Only 442 CVEs are attributable to Anthropic, about... og:url: https://daily.dev/posts/after-7-months-glasswing-overwelms-maintainers-with-underwhelming-findings-hljfurpi4 og:image: https://api.daily.dev/og/posts/HLJFURpi4.png og:image:alt: After 7 Months, Glasswing Overwelms Maintainers with Underwhelming Findings og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # After 7 Months, Glasswing Overwelms Maintainers with Underwhelming Findings **[Hacker News](https://daily.dev/sources/hn)** · 8 min read · 0 upvotes · 0 comments ## Summary Seven months into Anthropic's Project Glasswing, VulnCheck finds the results don't match the launch hype. Only 442 CVEs are attributable to Anthropic, about 0.5% of all CVEs published this year, and just two have been exploited in the wild. Of 37,966 reported findings, only 217 are marked fixed while 1,654 are duplicates and 419 withdrawn. curl's maintainer marked 19 of 34 published reports as not a vulnerability, confirming only six as low severity, putting the false positive rate at 55.9% so far against Anthropic's claimed 94.2% true positive rate. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### How many CVEs has Anthropic's Project Glasswing actually produced after seven months? 442 CVEs have been tracked as attributable to Anthropic seven months after Project Glasswing launched, an average of 63 per month. That is roughly 0.5% of the 93,450 CVEs published this year. Only two of those 442 CVEs have been confirmed exploited in the wild, a rate of just 0.45%. _Developers weighing AI vulnerability-scanning claims against real-world results can track this kind of analysis on daily.dev._ ### What percentage of Anthropic's reported vulnerability findings have actually been fixed? Only 217 of Anthropic's 37,966 reported findings have been marked fixed, a tiny fraction of the total. Just 7,349 (19.4%) of findings have even been made public. Since September, nearly 12,000 new findings were added but only 15 additional fixes were recorded, while duplicates jumped from 2 to 1,670. _Teams evaluating AI-driven security tooling claims can follow data like this on daily.dev before trusting the numbers._ ### How accurate were Anthropic's OSS Scanner vulnerability reports when curl reviewed them? curl maintainer Daniel Stenberg found that of 34 published reports from Anthropic's OSS Scanner, 19 were rejected as not vulnerabilities, 6 were confirmed as low severity, and 9 remained unassessed. That puts the false positive rate at 55.9%, with only 17.6% confirmed, all low severity, far below Anthropic's claimed 94.2% true positive rate. _Maintainers deciding whether to adopt AI vulnerability scanners can weigh real validation results shared on daily.dev._ ## Community take How the wider developer community reacted, aggregated from 2 discussions and 103 comments across hackernews (as of 2026-11-11). **TL;DR:** Most commenters see Project Glasswing's results as underwhelming and view the volume of mostly low-value findings as a burden shifted onto unpaid maintainers, though a vocal minority argues the analysis is misleading because it only covers public OSS disclosures, not proprietary use cases where results may be much better. **Sentiment:** 10% positive · 35% mixed · 55% skeptical **The case for** - One maintainer reported receiving several genuine, high-quality findings and fixes from Anthropic's tooling with no false positives in their case. - Some argue the methodology is flawed since it only analyzes the small, public slice of findings and can't speak to the private/proprietary deployments where fix rates are reportedly much higher. - A few note that finding vulnerabilities before they're exploited in the wild is the intended success case, not a failure, so low exploitation numbers shouldn't be read negatively. **The pushback** - Many see the piece's harsh framing as suspicious given it doubles as an advertisement for the author's own vulnerability-prioritization product. - Multiple people note the triage burden is overwhelming volunteer maintainers with large volumes of low-severity or non-bug reports. - Several point out Anthropic isn't providing adequate fixes alongside findings, and that proposed patches are sometimes low quality or themselves add noise. - Some compare the project to other AI-driven 'discovery' efforts (like DeepMind's materials project) where headline numbers mask that most 'findings' are duplicates or junk once verified by humans. **By community** - hackernews (heated): Discussion splits sharply between those convinced Glasswing's output is mostly noise burdening maintainers and those defending it as either effective in proprietary contexts or unfairly judged by a self-interested analysis. **Hottest debate:** Whether the high false-positive/low-fix rate reflects Glasswing's output being genuinely low quality, or whether the analysis unfairly judges it using only the small, skewed public subset of findings. **Open questions** - How do fix and accuracy rates compare between the public OSS disclosure program and Anthropic's use of the same models on proprietary codebases? - Why isn't Anthropic providing more usable patches alongside the vulnerability reports to reduce maintainer workload? - What is the real-world exploitation and harm rate from the unexploited vulnerabilities that remain undisclosed or unfixed? **Highlights** > I received 9 reports of Fil-C issues from Anthropic All of them were real bugs. None of them were false positives. The quality of the reports was exceptional. (My baseline is doing security defense on a browser engine, so my baseline has a large sample count and is probably representative of what experts are capable of.) The quality of the proposed fixes was alright. I took some of the fixes verbatim. I used most of the fixes as significant inspiration for doing it my way. The quality of the test cases was exceptional. Maybe I’m special somehow! But I just want to call out the enormous chasm between what I saw of anthropic’s capability and what TFA says > — [pizlonator on hackernews · 1 comments](https://news.ycombinator.com/item?id=50448745) > The blog is misleading because it tries to characterize Project Glasswing's effectiveness through use of a tangentially related data source. A Project Glasswing vulnerability discovery most often occurs when a Glasswing partner points a Mythos-class model at its own, often proprietary, software. The data source (aka "ledger") is only for the Anthropic coordinated vulnerability disclosure program for open source software: in other words, Anthropic itself pointing a Mythos-class model at an OSS package, and then notifying the maintainer afterward that it had done this, and had findings to share. This OSS work is certainly useful, but the number of fixes reflects - in part - limited time availability by volunteer OSS maintainers, and Anthropic sometimes working without knowledge of the OSS project's threat model (the threat model is explicitly requested in the newer oss-scanner effort). Percentage fixed is often much higher for use of Glasswing with proprietary software, e.g., https://blog.cloudflare.com/cyber-frontier-models/ "Our research with Mythos Preview was conducted in a controlled environment against our own code; every vulnerability surfaced through this work was triaged, validated, and remediated where action was needed under Cloudflare's formal vulnerability management process." > — [rurcliped on hackernews](https://news.ycombinator.com/item?id=50448662) > This is literally an ad for a service that helps you prioritize vulns! It is incredible how often an alleged study like this turns out to be an ad for such a service. "Businesses finding ROI hard to demonstrate" -> ad for a product that helps you demonstrate ROI. Etc. > VulnCheck is helping organizations not just to solve the vulnerability prioritization challenge - we’re working to help equip any product manager, CSIRT/PSIRT or SecOps team and Threat Hunting team to get faster and more accurate with infinite efficiency using VulnCheck solutions. > — [dcre on hackernews](https://news.ycombinator.com/item?id=50448780) > Maybe not. And that’s totally fine. But right now they’re not even offering one. They’re just putting the burden on the project without even trying to help fix it. They have all the resources in the world, and they won’t even try. And that’s of course assuming it’s a real bug, as we’ve seen a ton of them don’t seem to be. The last time we went through this, curl reported not only that a bunch of them weren’t bugs, but some of the bugs were listed previously on their mailing list, and that’s probably where the LLM got them. So not only is it finding small bugs. And nonbugs. It’s reporting previously found bugs. Pilling work on maintainers. Over and over again. To help? No. To make big numbers to show off for their IPO documents. > — [MBCook on hackernews](https://news.ycombinator.com/item?id=50448538) > > So far, only 2 of the 442 have been exploited in the wild, just 0.45%. > almost none are being exploited Isn't the whole point of glasswing to patch stuff way before it can be exploited? If the patched code is distributed quickly- there's no opportunity (or incentive) for bad actors to craft and distribute an exploit. This reads like an attempt to spin them as not being "real" vulnerabilities. > — [sigmar on hackernews · 1 comments](https://news.ycombinator.com/item?id=50448708) **Source threads** - [hackernews](https://news.ycombinator.com/item?id=50447074) · 231 points · 103 comments - [hackernews](https://news.ycombinator.com/item?id=50438624) · 2 points · 0 comments ## Similar posts on daily.dev - [Project Glasswing has uncovered 10,000 vulnerabilities: Anthropic](https://daily.dev/posts/project-glasswing-has-uncovered-10-000-vulnerabilities-anthropic-33ik2enrf) · CSO Online · 1 upvotes · 0 comments - [Behind the Mythos hype, Glasswing has just one confirmed CVE](https://daily.dev/posts/behind-the-mythos-hype-glasswing-has-just-one-confirmed-cve-km044wzwe) · CSO Online · 0 upvotes · 0 comments - [Mythos Vulnerability Firehose Hits a Human Bottleneck](https://daily.dev/posts/mythos-vulnerability-firehose-hits-a-human-bottleneck-rp6eq7jsa) · Dark Reading · 0 upvotes · 0 comments --- Tags: [#security](https://daily.dev/tags/security), [#open-source](https://daily.dev/tags/open-source), [#anthropic](https://daily.dev/tags/anthropic), [#ai-security](https://daily.dev/tags/ai-security), [#curl](https://daily.dev/tags/curl) [View this post on daily.dev](https://daily.dev/posts/after-7-months-glasswing-overwelms-maintainers-with-underwhelming-findings-hljfurpi4) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"After 7 Months, Glasswing Overwelms Maintainers with Underwhelming Findings","url":"https://daily.dev/posts/after-7-months-glasswing-overwelms-maintainers-with-underwhelming-findings-hljfurpi4","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/after-7-months-glasswing-overwelms-maintainers-with-underwhelming-findings-hljfurpi4"},"datePublished":"2026-11-10T23:59:06.761Z","dateModified":"2026-11-11T02:15:35.238Z","description":"Seven months into Anthropic's Project Glasswing, VulnCheck finds the results don't match the launch hype. Only 442 CVEs are attributable to Anthropic, about...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/efa3953e37acf7e251b32ad972b10a4e?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/efa3953e37acf7e251b32ad972b10a4e?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Hacker News","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Hacker News","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/hn","url":"https://daily.dev/sources/hn"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/after-7-months-glasswing-overwelms-maintainers-with-underwhelming-findings-hljfurpi4","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"security,open-source,anthropic,ai-security,curl","timeRequired":"PT8M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Hacker News","item":"https://daily.dev/sources/hn"},{"@type":"ListItem","position":3,"name":"After 7 Months, Glasswing Overwelms Maintainers with Underwhelming Findings"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/after-7-months-glasswing-overwelms-maintainers-with-underwhelming-findings-hljfurpi4#faq","mainEntity":[{"@type":"Question","name":"How many CVEs has Anthropic's Project Glasswing actually produced after seven months?","acceptedAnswer":{"@type":"Answer","text":"442 CVEs have been tracked as attributable to Anthropic seven months after Project Glasswing launched, an average of 63 per month. That is roughly 0.5% of the 93,450 CVEs published this year. Only two of those 442 CVEs have been confirmed exploited in the wild, a rate of just 0.45%. Developers weighing AI vulnerability-scanning claims against real-world results can track this kind of analysis on daily.dev."}},{"@type":"Question","name":"What percentage of Anthropic's reported vulnerability findings have actually been fixed?","acceptedAnswer":{"@type":"Answer","text":"Only 217 of Anthropic's 37,966 reported findings have been marked fixed, a tiny fraction of the total. Just 7,349 (19.4%) of findings have even been made public. Since September, nearly 12,000 new findings were added but only 15 additional fixes were recorded, while duplicates jumped from 2 to 1,670. Teams evaluating AI-driven security tooling claims can follow data like this on daily.dev before trusting the numbers."}},{"@type":"Question","name":"How accurate were Anthropic's OSS Scanner vulnerability reports when curl reviewed them?","acceptedAnswer":{"@type":"Answer","text":"curl maintainer Daniel Stenberg found that of 34 published reports from Anthropic's OSS Scanner, 19 were rejected as not vulnerabilities, 6 were confirmed as low severity, and 9 remained unassessed. That puts the false positive rate at 55.9%, with only 17.6% confirmed, all low severity, far below Anthropic's claimed 94.2% true positive rate. Maintainers deciding whether to adopt AI vulnerability scanners can weigh real validation results shared on daily.dev."}}]} ```