Skip to content
NewA calmer, faster ZeroHour

VulnCheck says Glasswing fixed 217 of nearly 38,000 claimed flaws

Original title: Anthropic's Glasswing Finds Many Vulnerabilities Yet Delivers Few Fixes, Researcher Says

Infosecurity Magazineby Kevin Poireault ()3 min read
Part of a story covered by 2 sourcesAnthropic Glasswing shows 217 fixes among 37,966 findingsMerged summary & timeline
AI summary · x-ai/grok-4.7Summaries by Vani AI

Of 5,276 validated ledger entries, Garrity says 217 are fixed, and he disputes Anthropic's 94.2% true-positive rate.

VulnCheck researcher Patrick Garrity says Anthropic's Project Glasswing has produced few public fixes seven months after launch. Anthropic's November 2 dashboard lists 37,966 findings, with 7,349 on the public ledger and 217 marked fixed. Anthropic has cited human triage limits, launched OSS Scanner, and sent nearly 5,000 unverified reports to maintainers who requested them. cURL maintainer Daniel Stenberg said many reports he reviewed were not vulnerabilities.

Unconfirmed: Garrity disputes Anthropic's 94.2% true-positive rate, and Anthropic had not commented by publication.

Key facts
  • Anthropic's November 2 ledger lists 7,349 of 37,966 claimed findings.
  • Garrity counts 5,276 unique validated issues, 1,654 duplicates and 419 withdrawals.
  • From September to November, 4,613 entries were added and 15 marked fixed.
  • Stenberg reviewed 34 of 45 cURL reports; 19 were not vulnerabilities.
  • Anthropic said October 8 it had manually reviewed about 6,000 of over 29,000 candidates.
Explain this AI news with Vani AIWhat was announced, what is new and why it matters, in a new chat

From the original article

3 min read at infosecurity-magazine.com

VulnCheck’s Patrick Garrity has found that seven months on, Project Glasswing resulted in only 217 vulnerabilities getting fixed, despite thousands reported

Infosecurity Magazine

By Kevin Poireault · infosecurity-magazine.com

Read the full article at Infosecurity Magazine

A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.