Latest / Neural Newscast / Cisco Patch Critical Flaws as McGraw Hill Leaks 13.5M Records [Prime Cyber Insights]
Transcript
- [00:00] Announcer: From Neural Newscast, this is Prime Cyber Insights, [00:03] Announcer: Intelligence for Defenders, Leaders, and Decision Makers. [00:06] Announcer: Welcome to Prime Cyber Insights for April 16th, 2026. [00:11] Announcer: We are moving fast today with a high-stakes infrastructure briefing. [00:15] Announcer: Joining me to analyze these developments is Lauren. [00:18] Aaron Cole: Thanks, Aaron. [00:19] Aaron Cole: We are starting with critical patches from Cisco. [00:22] Aaron Cole: They have disclosed four significant flaws in identity services and WebEx, [00:28] Aaron Cole: including several carrying a CVSS score of 9.9. [00:33] Aaron Cole: The exposure is severe, ranging from remote code execution to full user impersonation. [00:41] Announcer: Specifically, 2026 Cape Verde Escudos, 20,184, in WebEx. [00:49] Announcer: This involves improper certificate validation in the SSO integration, [00:53] Announcer: allowing unauthenticated remote attackers to impersonate any user. [00:58] Announcer: Simultaneously, we are tracking a massive leak from textbook giant McGraw-Hill. [01:03] Announcer: Reports from Have I Been Poned and the Register indicate a Salesforce-hosted page misconfiguration [01:10] Announcer: leaked 13.5 million records, including names and phone numbers. [01:15] Aaron Cole: While McGraw-Hill claims this did not touch their internal systems, the 100 gigabytes of data now appearing on dark web leak sites suggests a serious impact for the users involved. [01:28] Aaron Cole: Joining us today is Chad Thompson, a director-level security leader with a systems-level perspective on automation and risk. [01:37] Aaron Cole: Chad, we are seeing a new breed of vulnerabilities in AI-adjacent infrastructure, specifically [01:45] Aaron Cole: the engine UI bypassed involving the Model Context Protocol, or MCPages. [01:52] Aaron Cole: How is this changing the risk profile for standard web servers? [01:56] Lauren Mitchell: Lauren MCP is designed to allow AI agents to manage configurations directly [02:02] Lauren Mitchell: But in the case of NGXUI and CVE-2020-6032, it created an unauthenticated messages endpoint. [02:13] Lauren Mitchell: Attackers are using this to write malicious server blocks and reload configurations without any credentials. [02:21] Lauren Mitchell: It turns a management tool into a direct vector for total server takeover because the protocol assumptions did not match the application's security posture. [02:33] Announcer: Chad, look at the scale of this. [02:35] Announcer: Over 2,600 exposed instances found via Shodan. [02:40] Announcer: For a practitioner, is the lesson here that we should not be exposing management interfaces at all? [02:45] Announcer: Or is the integration of these AI protocols fundamentally flawed? [02:50] Lauren Mitchell: It comes down to shadow endpoints, Aaron. Developers often protect the initial session, [02:56] Lauren Mitchell: but leave command execution endpoints open to the network. In an automated agentic environment, [03:04] Lauren Mitchell: you must apply zero trust principles to every internal API call. [03:10] Lauren Mitchell: If your management UI supports a protocol like M-CP, [03:15] Lauren Mitchell: You must verify that the same Rbac and authentication headers apply to those new streaming endpoints. [03:24] Aaron Cole: We have seen similar logic gaps in the McGraw-Hill case, [03:28] Aaron Cole: where a limited Salesforce-hosted page led to millions of exposed records. [03:33] Aaron Cole: Chad, does this suggest we are moving into an era [03:36] Aaron Cole: where misconfigured integrations are more dangerous [03:40] Aaron Cole: than core software vulnerabilities themselves? [03:43] Lauren Mitchell: Absolutely, Lauren. [03:45] Lauren Mitchell: The core platforms like Salesforce or Nginx are often robust. [03:51] Lauren Mitchell: The breach happens in the glue. [03:53] Lauren Mitchell: the third-party plugins, and custom-hosted pages. [03:57] Lauren Mitchell: We are also seeing a rise in hijacked WordPress plugins, [04:01] Lauren Mitchell: where malicious actors buy legitimate plugin companies just to inject backdoors. [04:08] Lauren Mitchell: The system-level risk is now defined by the weakest link in your automated supply chain. [04:15] Announcer: Chad, thank you for that context. [04:17] Announcer: It frames why identity and integration points are the primary battleground right now. [04:23] Announcer: Lauren, this ties directly back to the UAC 24-7 activity reported by CERT-UA today, using AI-generated sites in compromised XSS vulnerabilities to deliver data-stealing malware to healthcare clinics in Ukraine. [04:39] Aaron Cole: It is a persistent theme. [04:42] Aaron Cole: Whether it is Cisco's SSO flaws, Salesforce misconfigurations, or malware targeting WhatsApp and Chromium browsers, the goal is always the identity. [04:54] Aaron Cole: For our listeners, the immediate move is clear. [04:58] Aaron Cole: Patch ISC to the latest versions, migrate to NGX UI 2.3.6, and audit any public-facing [05:07] Aaron Cole: Salesforce pages. [05:09] Announcer: We will be back tomorrow with more deep dives. [05:12] Announcer: For more details on these stories, visit pci.neuralnewscast.com. [05:17] Announcer: This podcast is a product of Neural Newscast, and the information provided is for educational [05:22] Announcer: purposes only. [05:24] Announcer: Always consult with your security team before making infrastructure changes. [05:28] Announcer: Neural Newscast is AI-assisted, human-reviewed. [05:32] Announcer: View our AI transparency policy at neuralnewscast.com. [05:37] Chad Thompson: This has been Prime Cyber Insights on Neural Newscast. [05:40] Chad Thompson: Intelligence for defenders, leaders, and decision makers.