Cybersecurity briefing, 2026-07-27

The AI agent didn't get hacked.

• Spring Boot fat-JARs stay exposed after Fastjson bypass
• AI agent became the intruder's hands-on operator
• Blockchain relays make botnet takedowns harder
• Windchill patch window became Cl0p's entry point
• AD CS trust check hands users Domain Controller power
• LockBit fell when its trust broke
• Heapdump scans can leak backend secrets
• Ransomware now kills EDR first
• Billing vendor breach spreads to seven providers
• Linux rootkits can lie to live-response tools
• Serv-U admin roles do not contain blast radius
• Crype service turns stealth into a commodity
• Public Wi‑Fi gateways become cross-tenant phishing points
• Dependency automation gets a time lock
• Dependabot slows the poisoned-release window
• Open AI security moves toward self-hosted control

Recent editions

• Sunday, July 26, 2026: Windchill patch window became Cl0p's entry point · 5 stories
• Saturday, July 25, 2026: Password reuse turns retail logins into fraud targets · 14 stories
• Friday, July 24, 2026: AI agents become intruders when access stays open · 15 stories
• Thursday, July 23, 2026: Langflow flaw puts connected systems at risk · 15 stories
• Wednesday, July 22, 2026: GitHub Actions turned into cPanel attack infrastructure · 17 stories
• Tuesday, July 21, 2026: Langflow intrusion now targets the model itself · 18 stories
• Monday, July 20, 2026: Overdue patch follows AI-found Sentry command injection · 17 stories