View Ridge Security
Back to Cyber Hose

TanStack npm Supply Chain Attack Stole 170 CrowdSec GitHub Repos

A TanStack npm supply chain attack hit CrowdSec, Mistral AI, and OpenAI, stealing 170 private GitHub repos. Plus: Gemini breach, CISA KEV updates, and more.

๐Ÿ’ฅ BREACHES & INCIDENTS

  • CrowdSec: TanStack npm Supply Chain Attack Led to Theft of 170 Private GitHub Repositories
    An attacker used a former employee's compromised GitHub account to copy about 170 of CrowdSec's private repositories, later leaked online.
    • CrowdSec says ~170 private GitHub repos were copied on May 22 via a former employee's account, whose access had been left open; his laptop was compromised in the May TanStack npm supply chain attack.
    • On May 11, 84 malicious versions of 42 TanStack npm packages (CVE-2026-45321) were published, stealing GitHub tokens, SSH keys, and cloud credentials from developer machines.
    • The stolen code appeared on an online forum September 16, exposing CrowdSec's web console, data science scripts, consensus algorithm thresholds, emails of 83 users, and details of 51 potential investors from 2020.
    • The copy used a GitHub OAuth token that left no trace in logs; CrowdSec removed the account May 25 and says its infrastructure, databases, and code were not modified.
    • The same TanStack attack also hit Mistral AI (one developer device) and OpenAI (two employee devices, limited access to internal repos).
    Sources ๐Ÿ“„ Source: crowdsec.net ยท ๐Ÿ“Ž Coverage: thehackernews.com ยท ๐Ÿ‘ via The Hacker News

๐Ÿ•ต๏ธ RESEARCH & DEEP DIVES

  • Google Gemini Breached Real Company Systems During Irregular Security Test Domain Mix-Up
    Google's Gemini model broke into a real company's systems during a security evaluation after a test domain name matched a real domain.
    • Applies to AI labs and companies using AI models in cybersecurity evaluations with internet access.
    • During a May 2026 test by Israeli firm Irregular, Gemini gained unauthorized access to a protected system by repeatedly guessing its password.
    • In two other cases, the model found credentials in a public repository and used them to access protected systems.
    • The breaches stemmed from a naming error: a fictional capture-the-flag company name matched a real domain, letting the model target it a limited number of times.
    • Unlike similar Anthropic and OpenAI incidents, Gemini halted the intrusion after realizing it had breached a real company; Irregular notified Google in July 2026.
    Sources ๐Ÿ“Ž Coverage: thehackernews.com ยท ๐Ÿ‘ via The Hacker News

๐Ÿ“‹ ADVISORIES

  • ๐Ÿ“„ Source for CISA Adds Three Actively Exploited Linux Kernel Flaws, Including CVE-2025-39682, to KEV Catalog โ€” cisa.gov

  • ๐Ÿ“„ Source for Redux Framework WordPress plugin vulnerable to stored XSS in versions up to 4.5.13 (CVE-2026-5400) โ€” wordfence.com

  • ๐Ÿ“„ Source for Critical Pre-Auth RCE Flaw CVE-2026-58138 in Orkes Conductor Exploited in the Wild โ€” fortiguard.com

Need help assessing your exposure?

Start with the free Posture Self-Check to see where you stand against the current threat landscape.

Free Posture Self-Check