Palo Alto, Cisco and Ivanti all patch actively exploited edge zero-days
Published 7:19 PM · 120 articles from 15 sources · about 5 min to read in full
Themes & analysis
Security appliances remain the attackers' favorite door
PAN-OS, Cisco Secure Email Gateway and Ivanti Connect Secure all had exploited zero-days disclosed together, and Ivanti exploitation dates back to October 1. Collabora/Nextcloud and GitLab add to the pattern of exposed, privileged services. Speculation: defenders may need to assume at least one perimeter product is compromised and invest in detection behind it.
AI is cutting the cost of finding and exploiting bugs
XBOW built a Chrome full-chain exploit in six days using a public model, Vghost was found with AI, and an AI agent found a critical OTP bypass on a production insurer site. Meanwhile Microsoft shipped a record 1,413 fixes. Speculation: patch volume and exploit speed are both rising, which makes prioritization by exploitation status more important than severity scores.
Trust in hardware and physical systems is breaking
Ledger confirmed an implant in wallet hardware, Danfoss refrigeration controllers were taken over in France to heat units to about 50C, and an OpenAI engineer faces spy charges after a recording device was found in a meeting room. Software patching does not address these. Supply-chain and physical-access assumptions deserve re-examination.
Agent security is the weakest layer in AI deployment
Many agent-related papers appeared today: browser agents approving unrequested access, poisoned Agent Skills, encrypted skill injection reaching 47.5% attack success, and several defenses such as Kairos. CSO Online argues the problem is labs skipping basics. Reconsider granting agents broad authority until confinement at the system boundary is standard.
Legal and financial pressure on AI is building
Eight states sued OpenAI, 40 attorneys general are watching Character.AI's bankruptcy sale, and analysts question whether AI infrastructure revenue can cover spending. ASML's strong orders point the other way. Speculation: liability and financing risk may constrain AI companies as much as technical limits.
Watch list
Ivanti has promised a separate privilege-escalation fix for Connect Secure. Watch for it and for further exploitation of CVE-2026-104935.
Watch for Ivanti's promised privilege-escalation patch
Ledger's investigation into tampered devices and the sales halt across seven Asian countries. Watch for the scope of affected units and attribution.
Cisco Secure Email Gateway and PAN-OS exploitation by unknown actors. Look for attribution, indicators of compromise and post-exploitation reports.
Watch for CISA KEV remediation deadlines for CVE-2026-0322 and CVE-2026-76519
The Character.AI Chapter 11 case, where 40 attorneys general are demanding data protections in any asset sale, including one to Disney.
Watch for Court filings on the asset sale
Time standards bodies will vote on letting UTC drift from Earth's rotation, which would end leap seconds.
Watch for The standards-body vote on the leap second proposal
State lawsuits against OpenAI over child safety. Watch OpenAI's response and whether more states join.