AI Radar

VulnCheck tracks vulnerabilities discovered by Anthropic

VulnCheck published a tracking report on vulnerabilities attributed to Anthropic's Project Glasswing, seven months after launch. VulnCheck claims few findings are public or fixed and questions Anthropic's true positive rate.

Who should care
Open-source maintainers and security teams handling AI-generated vulnerability reports.

Key facts

VulnCheck tracked 442 CVEs attributable to Anthropic, about 0.5% of CVEs this year
37,966 reported findings, 7,349 (19.4%) public, 217 marked fixed
About 12,000 findings added since September, with only 15 more fixes
1,654 findings (4.4%) marked duplicates, 419 withdrawn
Anthropic's claimed true positive rate rose from 90.8% to 94.2%
curl rejected 19 of 34 published reports as vulnerabilities

Context

Project Glasswing launched on April 7th, per the VulnCheck post.

Timeline

Discussion

  1. Tracking Anthropic-discovered vulnerabilitiesvulncheck.com · HN · 447 pts · 217 comments

Evidence

Low evidence: a single community post. Treat it as a claim until someone else confirms it.

Details written Nov 11, 11:26 UTC from the sources listed above.