daily2026-11-20Covers → x-ai/grok-4.7Summaries by Vani AI9 · 3 reads · https://zerohour.day/brief/2026-11-20
Top story
Broadcom patches exploited VMware vCenter zero-day. Mandiant says Armory Castle used CVE-2026-123890 for administrator SSO sessions and stole vCenter directory databases. Network access alone can control managed virtual machines, and Broadcom lists no official workaround.
Exploitation & threats
- CISA lists exploited Synology, LiteLLM, Axis and REDCap flaws, including remote code execution, after finding active exploitation on November 18.
- Exploited Puppet Enterprise RCE added to CISA KEV: any Puppet CA-signed client can run code as the server service user through CVE-2026-115731.
- Attackers exploit Cisco Expressway flaw after public PoC; F6 says intruders rooted clustered boxes, while Cisco’s advisory still reports no malicious use.
- Attackers exploit PHP2Shell on exposed PAN-OS interfaces to gain root; Palo Alto confirmed unauthenticated code execution, and a chained kernel flaw yielded root on 24 firewalls.
- Actively exploited Apache Roller flaw allows unauthenticated remote code execution on versions 4.0.0 through 6.1.4 via XML-RPC deserialization before authentication.
- Malicious Microsoft agent-framework PyPI releases steal cloud credentials; StepSecurity, OX Security and Socket say versions 1.26.1 and 1.26.2 run a startup hook.
Patch priorities
- CVE-2026-123890, vCenter: upgrade to 9.1.1.0200, 9.0.2.0200, or 8.0 U3m using the standard update, not Quick Patch.
- CVE-2026-115731, Puppet Enterprise: upgrade to 2023.8.13, 2025.12.1, or 2026.0.1 and restart the server.
- CVE-2026-129448, Apache Roller: upgrade to 6.1.6 or later; 6.1.5 left the bug unfixed.
- CVE-2026-76466, Expressway: upgrade clustered systems to X15.5.20.
- CVE-2026-123007, nginx: upgrade to 1.30.7 or 1.31.9 and fully restart; Cloudflare reports no in-the-wild exploitation.
- CVE-2026-126714, PHP: update to 8.2.36, 8.3.37, 8.4.28, 8.5.13, or 8.6.0.
Breaches & incidents
- Hackers steal data of 1.6 million SBC Medical Group customers; an SEC filing says Japan databases exposed 1,624,000 people, some passwords, and 459 employees’ bank details.
- European Commission confirms October cyberattack involving an AI tool; personnel systems were reportedly offline more than 10 days, with no confirmed broad staff-data breach or ransom.
- Leak sites claimed Boeing, Eni, Smith & Wesson, UCHealth, Frontier Communications, and AutoZone, among others; those posts are group claims, not confirmed compromises.
AI
- Anthropic prices $81.3 billion IPO above SpaceX's record raise; shares priced at $83 begin Nasdaq trading Thursday under ANT, with a fully diluted value near $1.5 trillion.
- US export rules freeze China AI chip shipments and overseas cloud access; BIS licensing now covers overseas clusters of about 300 AI chips, with a 90-day transition.
- Google DeepMind releases Gemini 4 Neon, its fourth model in seven weeks, with a one-million-token window for text, images, audio, and video.
- Anthropic CEO recommits to pacing AI at Nasdaq debut; Amodei said training and deployment will wait until systems are expected to stay under human control.
Watchlist
- Federal agencies must fix Synology DSM, Apache Axis, and REDCap by November 21, LiteLLM by December 2, and Puppet Enterprise by December 1.
- vCenter 7.0 has no public fix for CVE-2026-123890; extended-support customers should contact Broadcom Support.
- Attackers probe Java4Shell; sensors have not confirmed a successful intrusion, and the bug needs an AVX-512 VBMI CPU.
- A Linux SCTP use-after-free tied to nginx2root was unfixed as of November 18, and Veria Labs plans a Ruby2Shell proof of concept within a week.
Written with Vani AI, the AI workspace from AhuraSense Ltd, the team behind ZeroHour. Free to try, no account needed.