{"slug": "vulncheck-glasswing-november-update-442-cves-217-fixes-curl-triage", "title": "VulnCheck Glasswing November Update: 442 CVEs, 217 Fixes, curl Triage", "summary": "VulnCheck researcher Patrick Garrity's November 9, 2026 update on Anthropic's Project Glasswing found 442 CVEs credited to Anthropic, about 0.5% of roughly 93,450 CVEs filed in 2026, with only 2 of the 442 (about 0.45%) matching Known Exploited Vulnerabilities. Anthropic's November 2 disclosure ledger showed 37,966 total findings against just 217 fixes, with duplicates rising from 2 to 1,670 and withdrawals from 245 to 419 since September 8, while Anthropic's claimed true positive rate climbed from 91.4% to 94.2%. Garrity's analysis, which also covers Daniel Stenberg's triage of Anthropic OSS Scanner reports sent to curl (19 of 34 partial results judged not vulnerabilities, 6 low, 9 pending), argues the volume-fix-severity gap still outruns maintainable remediation.", "body_md": "Seven months after [Project Glasswing](https://explainx.ai/blog/claude-mythos-preview-cybersecurity-glasswing) launched, [VulnCheck researcher Patrick Garrity](https://www.vulncheck.com/blog/anthropic-glasswing-november-update?ref=explainx) published a **November 9, 2026** follow-up that pairs **Anthropic’s public disclosure ledger** with **CVE attribution**, **Known Exploited Vulnerability (KEV)** matches, and **Daniel Stenberg’s live triage** of [Anthropic OSS Scanner](https://explainx.ai/blog/anthropic-oss-scanner-free-ai-vulnerability-scans-open-source-maintainers-2026) reports sent to **curl**. The piece does not claim frontier cyber models are useless; it argues the **volume–fix–severity story** still outruns maintainable remediation.\n\nThe analysis extends Garrity’s [July exploitation report](https://explainx.ai/blog/vulncheck-ai-found-bugs-exploit-rate-glasswing-july-2026), which found AI-assisted public CVEs exploited at roughly the **baseline KEV rate**. This update adds **ledger mechanics**: surging **duplicates and withdrawals**, **stalled fixes** while findings grow, and **maintainer-side validation** after Anthropic shifted unreviewed batches to projects that opt in.\n\n## TL;DR\n\n| Question | Answer per VulnCheck (Nov 9, 2026) | \n|---|---|\n| CVEs credited to Anthropic? | **442** tracked in Garrity’s repo | \n| Share of 2026 CVEs? | About **0.5%** of ~93,450 | \n| Ledger findings / fixed? | **37,966** /**217** on Nov 2 ledger | \n| Public on ledger? | **7,349 (19.4%)** | \n| Known exploited in KEV? | **2** of 442 (~0.45%) | \n| curl OSS Scanner triage (partial)? | **19/34** not vulns;**6** low;**9** pending | \n| Anthropic claimed true positive rate? | **94.2%** on ledger vs earlier**90.8%** | \n\nAgents that act on your behalf need rules. [See how explainx helps secure your agents →](https://explainx.ai/agentbeam?src=blog_note)\n\n## What changed since the April Glasswing launch\n\nAnthropic announced [Project Glasswing on April 7, 2026](https://www.anthropic.com/news/project-glasswing?ref=explainx), saying partners had already found **thousands of high- and critical-severity vulnerabilities** heading toward coordinated disclosure. That headline set expectations for **maintainers**, **investors**, and **lawmakers** debating [Mythos-class access](https://explainx.ai/blog/gpt-6-cyber-vs-claude-mythos-5-glasswing-claude-security-comparison-2026).\n\nGarrity began an independent ledger on **April 17** in the open [**Anthropic Credited CVEs**](https://github.com/vulncheck-oss/anthropic-credited-cves?ref=explainx) repository because Anthropic’s own [** disclosure ledger**](https://www.anthropic.com/transparency/vulnerability-disclosure-ledger?ref=explainx) updates in **infrequent bulk batches**. His November post is the fifth in a series; our [July coverage](https://explainx.ai/blog/vulncheck-ai-found-bugs-exploit-rate-glasswing-july-2026) already compared **AI discovery counts** to **KEV exploitation share**.\n\nThe new numbers focus on **remediation throughput** and **severity alignment**, not whether a model can spot a bug in a lab.\n\n## Ledger growth versus fixes and public disclosure\n\nAnthropic’s **November 2** ledger snapshot, as summarized by Garrity:\n\n| Metric | Sep 8 | Nov 2 | Change | \n|---|---|---|---|\n| Total findings | 26,153 | 37,966 | +11,813 | \n| Made public | 2,736 (10.5%) | 7,349 (19.4%) | +4,613 | \n| Fixed | 202 | 217 | +15 | \n| Withdrawn | 245 | 419 | +174 | \n| Duplicates | 2 | 1,670 | +1,668 | \n| Claimed true positive rate | 91.4% | 94.2% | +2.8 pts | \n\nGarrity notes **nearly 12,000 findings** arrived in roughly two months while only **15** additional fixes registered. **Duplicates** jumped from **two to thousands**, matching his September warning that unresolved queues collide with human discovery elsewhere.\n\nHe also flags **1,654 duplicates (4.4%)** and **419 withdrawals** against **217 fixes**, and questions how a rising **94.2% true positive** rate squares with withdrawals and duplicates piling up faster than remediated items.\n\nProject Glasswing November ledger totals tracked by VulnCheck for Anthropic vulnerability findings\n*Image: [VulnCheck](https://www.vulncheck.com/blog/anthropic-glasswing-november-update?ref=explainx)*\n\n**Builder takeaway:** if you ingest Glasswing-style feeds into CI, budget **deduplication** and **severity re-ranking** pipelines, not just ticket openers. Our [Trail of Bits patching regression work](https://explainx.ai/blog/trail-of-bits-patching-bugs-regression-hunt-plugins-21-regressions-130-prs-patch-the-planet-2026) shows fix quality matters as much as find volume.\n\n## CVE attribution versus ledger marketing\n\nOn the CVE side, Garrity reports **442** vulnerabilities attributable to Anthropic in his tracker, **more than double** the **217** fixes Anthropic’s ledger lists. That gap can reflect **disclosure timing**, **non-CVE advisories**, and **partner-private work**, but it is the reference set VulnCheck uses for **exploitation math**.\n\nAgainst **VulnCheck KEV**, only **two** credited CVEs show **known exploitation**:\n\n- **CVE-2026-61500**\n- **CVE-2026-26980** (also discussed in our[July Glasswing exploitation post](https://explainx.ai/blog/vulncheck-ai-found-bugs-exploit-rate-glasswing-july-2026) )\n\nThat is **0.45%** of the 442-set. Garrity contrasts the **Mythos-era danger narrative** with **public exploit telemetry**, while stressing AI still shrinks **time-to-patch** when triage works.\n\nReaders comparing vendors should cross-check [OpenAI GPT-6 Cyber exploit pipelines](https://explainx.ai/blog/openai-gpt-6-cyber-critical-exploitgym-honeypot-under-1-percent-out-of-bounds-daybreak-red-2026), [Collinear CWE-bench defensive tasks](https://explainx.ai/blog/collinear-cwe-bench-v2-100-defensive-patch-tasks-91-cwes-gemini-4-argon-67-8-percent-2026), and [UK AISI cyber eval incidents](https://explainx.ai/blog/aisi-mythos-5-gpt-5-6-sol-cyber-eval-incident-august-2026) to see **offense-heavy marketing** versus **patch-loop benchmarks**.\n\n## OSS Scanner: triage moved to maintainers\n\nWhen human review could not scale, Anthropic launched the [**OSS Scanner**](https://www.anthropic.com/news/open-source-vulnerability-scanner?ref=explainx) on **October 8, 2026**, noting **29,000+ candidate findings** but only **~6,000** manually reviewed, with **~5,000 unverified reports** already emailed to maintainers who requested bulk delivery. Reports are **fully model-generated without human triage**, per Anthropic’s disclaimer.\n\nGarrity’s framing: the **bottleneck did not disappear**; it **moved downstream** to volunteers and corporate maintainers already stretched by [PHP2Shell response](https://explainx.ai/blog/php2shell-cve-2026-109618-109619-what-happened-what-to-do-now-2026) and routine CVE load.\n\n## curl gives a ground-truth sample\n\nThree days before the VulnCheck post, **curl maintainer Daniel Stenberg** described receiving **45 vulnerability reports** from Anthropic’s scanner path, rating **one medium** and the rest **low**, and publishing issues **slowly** to avoid drowning the team.\n\nOf **34 reports published** when Garrity wrote:\n\n| Outcome | Count | \n|---|---|\n| Not a vulnerability | 19 | \n| Confirmed low severity | 6 | \n| Awaiting assessment | 9 | \n\nThat yields a **55.9% false-positive rate** on published items so far and **17.6% confirmed**, all **low** severity in the confirmed bucket. Stenberg linked the batch on his blog; Garrity embeds VulnCheck’s chart.\n\ncurl maintainer triage chart for Anthropic OSS Scanner Glasswing vulnerability reports\n*Image: [VulnCheck](https://www.vulncheck.com/blog/anthropic-glasswing-november-update?ref=explainx)*\n\nThis rhymes with Stenberg’s **May** note that **five Mythos findings** collapsed to **one** confirmed curl issue, and with Garrity’s **September** severity mismatch: Anthropic labeled **91.5%** of ledger items critical or high while maintainers rated **51.3%** that way in his earlier sample.\n\nIf you maintain open source, treat scanner email like **untrusted fuzzer output**: require **reachable exploit paths**, **default-config impact**, and **regression tests** before issuing CVEs. Our [zero-day explainer](https://explainx.ai/blog/what-is-a-zero-day-vulnerability-exploit-explained-ai-era-2026) walks disclosure clocks; OSS Scanner explicitly **does not** run the same coordination timeline.\n\n## What developers are saying\n\nHacker News readers argued past the headline in both directions on [thread 50447074](https://news.ycombinator.com/item?id=50447074&ref=explainx):\n\n- **sigmar** ([comment 50448708](https://news.ycombinator.com/item?id=50448708&ref=explainx) ) said fast patching can keep KEV counts low, so**low exploitation share is not proof of failure** .\n- **pizlonator** ([comment 50448745](https://news.ycombinator.com/item?id=50448745&ref=explainx) ) reported**nine Fil-C reports from Anthropic** , all**real bugs** with**exceptional report quality** , contrasting curl’s batch.\n- **simoncion** ([comment 50449972](https://news.ycombinator.com/item?id=50449972&ref=explainx) ) pointed to**Greg Kroah-Hartman’s** high false-positive anecdotes and argued commenters ignored**OSS Scanner stats** while debating ledger percentages.\n- **rurcliped** ([comment 50448662](https://news.ycombinator.com/item?id=50448662&ref=explainx) ) said the post**mischaracterizes Glasswing** because many discoveries happen on**private partner code** never destined for the public ledger.\n- **enthuseyourcurb** ([comment 50448590](https://news.ycombinator.com/item?id=50448590&ref=explainx) ), claiming big-tech experience, argued**public CVE under-counts** company-ending issues that never receive numbers.\n- **delichon** ([comment 50447517](https://news.ycombinator.com/item?id=50447517&ref=explainx) ) worried about a**find-versus-fix ratio** of**37,966 to 217** if it generalizes.\n\nNone of these comments are verified facts; they show **maintainer burden**, **metric selection**, and **private versus public discovery** as the live debate.\n\n## What this means for agent and AppSec teams\n\n**Discovery tools are not patch tools.** If you wire [Claude Security](https://explainx.ai/blog/claude-security-scan-fix-public-beta-2026) or third-party scanners into repos:\n\n1. **Cap auto-ticket rate** and require**human or policy gate** for severity above low.\n2. **Deduplicate against existing GHSA/CVE** before paging on-call.\n3. **Measure time-to-merge fix PR** , not raw finding count OKRs.\n4. **Separate partner-private red team results** from**public ledger KPIs** when reporting upward.\n\nFor **agent pipelines** that write patches, pair scanning with **regression harnesses** and runtime guardrails. [AgentBeam](https://agentbeam.com/?ref=explainx) focuses on **stopping autonomous agents** from executing dangerous actions when scanners or coders go haywire; it complements, not replaces, maintainer triage.\n\n## How to verify the numbers yourself\n\n1. Clone or browse [**vulncheck-oss/anthropic-credited-cves**](https://github.com/vulncheck-oss/anthropic-credited-cves?ref=explainx) and diff against[Anthropic’s ledger CSV exports](https://www.anthropic.com/transparency/vulnerability-disclosure-ledger?ref=explainx) .\n2. Filter the ledger for **disclosed** versus**fixed** and compare Garrity’s claim about**disclosed inflation** using Anthropic’s UI filters.\n3. Watch **curl GitHub issues** filed from the Anthropic batch for final dispositions.\n4. Cross-check **KEV** entries for CVE-2026-61500 and CVE-2026-26980 in[CISA’s catalog](https://www.cisa.gov/known-exploited-vulnerabilities-catalog?ref=explainx) .\n\nAnthropic has not published a point-by-point rebuttal to this VulnCheck post in the materials we fetched on **November 11, 2026**. Watch its transparency pages and [Mythos 5.2 security scans](https://explainx.ai/blog/claude-mythos-5-security-scans-defender-advantage-fund-august-2026) for counter-metrics.\n\n## Why public CVE math understates partner work — and why it still matters\n\nGlasswing’s public ledger is only the **coordinated disclosure** slice. Anthropic has repeatedly said **Mythos-class partners** scan **proprietary stacks** that may never become CVEs. That is real defender value, and HN commenters such as **enthuseyourcurb** argued corporate incidents will never appear in Garrity’s CSV.\n\nThe tension is **accountability**. When Anthropic’s April launch quoted **thousands of critical findings** heading to open-source maintainers, policymakers and CISOs reasonably watched **fix velocity** on **public code**. Seven months later, **217 fixes** on **37,966 ledger rows** is a signal about **operational scaling**, even if private partner counts look better.\n\nFor teams buying **Claude Enterprise security scans**, ask vendors for **your** metrics: **true positive rate after human review**, **median days to merged fix**, **severity after maintainer downgrade**, and **duplicate rate against existing GHSA**. Do not import ledger marketing KPIs into your SOC without that filter.\n\nComparing offense models, [Vals AI cyber range results](https://explainx.ai/blog/vals-ai-cyberoffense-gpt-6-sol-daybreak-deepseek-v4-1-flash-open-weight-cyber-range-2026) show how **gating** shapes measured risk versus open weights. Glasswing’s structure keeps Mythos behind partners; OSS Scanner pushes **ungated model output** straight to inboxes. curl’s triage is the first visible **quality control point** on that new path.\n\n## Related reading\n\n*Ledger figures, curl triage counts, and KEV matches reflect VulnCheck’s November 9, 2026 post and linked maintainer blogs as of this article’s publication date.*", "url": "https://wpnews.pro/news/vulncheck-glasswing-november-update-442-cves-217-fixes-curl-triage", "canonical_source": "https://explainx.ai/blog/vulncheck-glasswing-november-ledger-curl-oss-scanner-2026?utm_source=rss&utm_medium=feed&utm_campaign=blog", "published_at": "2026-11-11 00:00:00+00:00", "updated_at": "2026-11-11 07:20:13.850050+00:00", "lang": "en", "topics": ["ai-safety", "ai-policy", "artificial-intelligence"], "entities": ["VulnCheck", "Anthropic", "Patrick Garrity", "Project Glasswing", "Daniel Stenberg", "curl", "Anthropic OSS Scanner", "Known Exploited Vulnerability"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/vulncheck-glasswing-november-update-442-cves-217-fixes-curl-triage", "markdown": "https://wpnews.pro/news/vulncheck-glasswing-november-update-442-cves-217-fixes-curl-triage.md", "text": "https://wpnews.pro/news/vulncheck-glasswing-november-update-442-cves-217-fixes-curl-triage.txt", "jsonld": "https://wpnews.pro/news/vulncheck-glasswing-november-update-442-cves-217-fixes-curl-triage.jsonld"}}