ON 12 October 2026, the SANS Internet Storm Center’s Stormcast covered a mix of threat intelligence governance and concrete vulnerability snapshots. The host clarifies the Traffic Light Protocol’s purpose: it indicates who information can be shared with, but not how it should be shared or safeguarded. This distinction is emphasised as a reminder that “amber” sharing still requires appropriate protective measures, rather than blanket encryption or internal classifications.
The discussion then pivots to a concrete vulnerability in Amazon’s Bedrock AI agents, where server-side request forgery (SSRF) was demonstrated. Researchers from Zenity Labs showed that a script running inside an EC2 VM could access the instance metadata service to harvest credentials, which an attacker could abuse. The root issues highlighted are the continued use of IMDS version 1 instead of version 2 and overly broad privileges that permissions granted to the agent did not require.
Amazon has since fixed the flaw, but the takeaway remains: when building agents or similar services, restrict URL access and favour IMDSv2 for hardened SSRF resistance.
The episode also notes a run of NetScaler (Citrix) vulnerabilities and stresses the importance of timely patching, given the tendency for ransomware operators to exploit exposed VPN surfaces. ReliaQuest is cited as observing ransomware deployed via PAN-OS vulnerabilities affecting GlobalProtect and Prisma Access. The host urges diligent patch management for VPN endpoints and client updates, and flags ongoing privilege-escalation issues in VPN clients as an ongoing concern.
Overall, the broadcast ties governance and practical exposure together—update the infrastructure, restrict agent capabilities, and remain vigilant for remote-access entry points used by threat actors.