AI Needs Managers Now? | Smart Glasses Return & Mythos Finds 23,000 Bugs

AI Needs Managers Now? | Smart Glasses Return & Mythos Finds 23,000 Bugs

In this episode of IT SPARC Cast - News Bytes, John & Lou explore how AI is rapidly evolving from simple assistants into autonomous workers that require management, oversight, and governance. Google introduces an open-source Agent Executor framework designed to supervise AI agents in production environments, while smart glasses may finally be approaching the point where they become practical for mainstream use. The episode also dives into the growing impact of AI-driven cybersecurity. Anthropic’s Mythos platform identified more than 23,000 potential vulnerabilities across open-source projects, raising important questions about how the industry will keep pace with validation, patching, and deployment. If you work in enterprise IT, cloud, cybersecurity, or AI, this episode offers a glimpse into where the next wave of operational challenges is headed.   ⸻ 📌 Show Notes 00:00 – Intro This week’s episode covers AI agent management, the future of smart glasses, and the growing challenge of handling AI-discovered software vulnerabilities. ⸻ 📰 News Bytes 00:48 – Google Adds Open Source Agent Executor Google announced an open-source Agent Executor framework designed to help organizations safely run AI agents in production. The platform provides orchestration, task management, state tracking, auditing, and recovery workflows for fleets of AI agents. John & Lou compare the concept to middle management for AI—providing oversight, accountability, and guardrails that help prevent autonomous systems from making costly mistakes. Key takeaways: * AI agents require supervision and governance * Enterprises need auditing and recovery mechanisms * Agent fleets will require dedicated management infrastructure https://www.computerworld.com/article/4176809/google-adds-open-source-agent-executor-to-support-ai-agents-in-production-3.html ⸻ 07:19 – Smart Glasses: Are They Getting Real? XREAL and Google continue pushing augmented reality forward with new Android XR initiatives and lightweight smart glasses designs. Improvements in AI assistants, displays, optics, and battery technology are bringing wearable computing closer to practical adoption. The discussion explores whether smart glasses are finally approaching an inflection point where they move beyond niche devices and become a true successor—or companion—to smartphones. Key considerations: * AI assistants significantly increase utility * Wearables face challenges around battery life and social acceptance * AR development platforms may become the next major ecosystem battle https://techcrunch.com/2026/05/24/xreal-googles-smartglasses-partner-thinks-it-has-finally-mastered-this-notoriously-tricky-industry/ ⸻ 14:31 – Mythos Detected 23,000 Potential Vulnerabilities Anthropic revealed that its Mythos platform identified more than 23,000 potential vulnerabilities across approximately 1,000 open-source projects during limited testing. Over 1,700 findings were independently validated, including more than 1,000 high or critical severity issues. While AI is dramatically accelerating vulnerability discovery, the larger challenge may now be validation, patching, distribution, and deployment. Finding the bugs is no longer the bottleneck. Key takeaways: * AI is transforming vulnerability research * Patching and deployment remain major obstacles * Open-source communities may need new funding and workflow models https://www.securityweek.com/anthropic-mythos-detected-23000-potential-vulnerabilities-across-1000-oss-projects/ ⸻ 🔚 20:49 – Wrap Up As AI systems become more autonomous, organizations must rethink how they manage software development, cybersecurity, and operational governance. The future may belong not just to AI tools, but to the frameworks that supervise them safely and effectively. ⸻ 🌐 Social Links IT SPARC Cast @ITSPARCCast on X https://www.linkedin.com/company/sparc-sales/ on LinkedIn John Barger @john_Video on X https://www.linkedin.com/in/johnbarger/ on LinkedIn Lou Schmidt @loudoggeek on X https://www.linkedin.com/in/louis-schmidt-b102446/ on LinkedIn ---------------------------------------- Hosted on Acast. See acast.com/privacy [https://acast.com/privacy] for more information.

Underminr Explained: The CDN Attack That Hides Malware Behind Trusted Traffic

A newly disclosed attack technique called “Underminr” allows malicious traffic to hide behind trusted CDN infrastructure, potentially bypassing DNS filtering, zero trust policies, and traditional security controls. In this episode of IT SPARC Cast – CVE of the Week, John and Lou explain how attackers abuse TLS routing and CDN tenant behavior to disguise command-and-control traffic as legitimate web traffic — and why AI-driven behavioral analysis may become the only effective defense. ⸻ 📄 Show Notes 🚨 CVE of the Week: Underminr This week’s episode focuses on Underminr, a stealthy attack technique that allows malicious traffic to hide behind trusted CDN infrastructure. The attack abuses: * CDN tenant routing * TLS SNI mismatches * HTTP host header manipulation * DNS resolution inconsistencies The result: Malicious command-and-control traffic can appear to originate from trusted services such as CDN providers. ⸻ ⚠️ Why This Is Dangerous Traditional security controls often trust: * Well-known domains * CDN traffic * TLS-encrypted connections Underminr exploits that trust model. Potential impacts include: * Bypassing DNS filtering * Evading protective DNS systems * Hiding malware communications * Concealing data exfiltration * Circumventing outbound filtering policies Because CDNs naturally move large volumes of traffic, malicious transfers can blend into legitimate content distribution activity. ⸻ 🛠️ Mitigation Steps for Underminr ✅ Validate TLS and Routing Consistency Verify that: * DNS resolution * TLS SNI fields * HTTP host headers * CDN routing destinations …all match expected destinations. This is one of the most important defenses. ⸻ ✅ Implement Deep Packet Inspection (DPI) Traditional DNS filtering alone is no longer enough. Use: * TLS inspection * Deep packet inspection * Proxy inspection * Behavioral traffic analysis to identify suspicious traffic patterns. ⸻ ✅ Deploy Behavioral Network Analytics Monitor for: * Unusual CDN usage * Unexpected outbound transfers * Off-hours synchronization activity * Abnormal traffic paths Example: A large CDN upload occurring at 3AM outside normal workflows should trigger investigation. ⸻ ✅ Enforce Zero Trust Outbound Policies Instead of trusting domains: * Validate applications and processes * Restrict outbound communication permissions * Use application-aware filtering * Limit which services can communicate externally ⸻ ✅ Improve CDN Isolation Policies CDN providers should: * Tighten tenant routing validation * Prevent cross-tenant hostname abuse * Restrict mismatched origin routing ⸻ 🤖 AI and the Future of Network Security John and Lou discuss how AI-assisted security analytics may become essential against attacks like Underminr. Traditional rule-based systems struggle with: * Correlating multiple protocol layers * Detecting subtle routing anomalies * Identifying behavioral inconsistencies in real time AI-driven network analysis could help identify: * Suspicious traffic paths * Out-of-sequence synchronization * Unusual CDN behavior * Hidden command-and-control channels ⸻ 💬 Listener Feedback Thanks to listeners Ahmed and Dennis for the feedback on last week’s Exchange vulnerability episode. One major takeaway: Organizations continuing to run on-prem email infrastructure are increasingly carrying significant operational and security risk. ⸻ 📣 Wrap Up Do you think traditional network trust models are finally breaking down, or can modern AI-driven security tools adapt quickly enough? 📧 feedback@itsparccast.com 🐦 @itsparccast on X ⸻ 🔗 Social Links IT SPARC Cast @ITSPARCCast on X https://www.linkedin.com/company/sparc-sales/ on LinkedIn John Barger @john_Video on X https://www.linkedin.com/in/johnbarger/ on LinkedIn Lou Schmidt @loudoggeek on X https://www.linkedin.com/in/louis-schmidt-b102446/ on LinkedIn ---------------------------------------- Hosted on Acast. See acast.com/privacy [https://acast.com/privacy] for more information.

AI Data Centers, Vibe-Coded Android Apps, and the Coming Security Flood

In this episode of IT SPARC Cast - News Bytes, John & Lou break down the growing AI infrastructure arms race, Google’s push toward AI-generated mobile apps, and Cloudflare’s latest findings on frontier AI security models. As AI compute demand explodes, the conversation explores how power generation, cloud scaling, and automation are rapidly reshaping enterprise IT. They also dive into Google’s new AI Studio tools that let users build Android apps in minutes and discuss how AI-driven vulnerability research is changing software security forever. From multi-agent bug hunting systems to the future of software development itself, this episode looks at how AI is transforming both how software is built—and how it’s secured. ⸻ 📌 Show Notes 00:00 – Intro ⸻ 📰 News Bytes 00:49 – xAI Power, Anthropic Workloads Anthropic signed a massive compute deal with SpaceX/xAI worth potentially tens of billions of dollars, highlighting how compute capacity has become the primary bottleneck in AI growth. The discussion explores the rise of “Neo Cloud” providers, AI-driven data center expansion, and the enormous power requirements driving demand for natural gas, nuclear energy, and eventually orbital data centers. Key takeaways: * AI revenue is increasingly tied directly to compute availability * Data center power generation is becoming a strategic industry * SpaceX and xAI are positioning themselves as major AI infrastructure providers https://techcrunch.com/2026/05/20/anthropic-will-pay-xai-1-25-billion-per-month-for-compute/ https://techcrunch.com/2026/05/20/musks-xai-is-being-sued-over-its-data-center-generators-now-its-buying-2-8b-more/ ⸻ 06:48 – Google’s AI Studio Lets Anyone Build Android Apps Google announced major upgrades to AI Studio that allow users to generate Android apps directly from text prompts using AI. The tools support hardware integrations like GPS, Bluetooth, and NFC while enabling users to preview and export apps quickly. John & Lou discuss how this may shift app development away from traditional coding and toward personalized automation and workflow control. Key considerations: * AI-assisted development dramatically lowers barriers to entry * App development may become more task-oriented than platform-oriented * Security and app validation remain major concerns https://techcrunch.com/2026/05/19/googles-ai-studio-now-lets-anyone-build-android-apps-in-minutes/ [https://techcrunch.com/2026/05/19/googles-ai-studio-now-lets-anyone-build-android-apps-in-minutes/] ⸻ 12:11 – Cloudflare Reports on Frontier AI Models & Security Cloudflare published findings from Project Glasswing and Anthropic’s Mythos model, revealing major advances in AI-driven vulnerability discovery. The report shows how specialized AI models can now identify exploit chains, generate proofs of concept, and assist with patch validation far beyond traditional coding agents. However, false positives, prompt bypasses, and scaling issues remain significant challenges. Key takeaways: * AI vulnerability hunting is advancing rapidly * Generic coding agents struggle with deep security analysis * Software architecture and patching workflows must evolve for the AI era https://blog.cloudflare.com/cyber-frontier-models/ ⸻ 📬 25:16 – Mail Bag Listener feedback highlights growing excitement around the show’s AI coverage and sparks additional discussion around the future of security operations, AI-assisted coding, and enterprise infrastructure strategy. 🔚 26:02 – Wrap Up As AI accelerates software development and vulnerability discovery simultaneously, enterprise IT teams will need stronger architecture, better automation, and tighter security discipline than ever before. The future isn’t just AI-powered—it’s AI-amplified. ⸻ 🌐 Social Links IT SPARC Cast @ITSPARCCast on X https://www.linkedin.com/company/sparc-sales/ on LinkedIn John Barger @john_Video on X https://www.linkedin.com/in/johnbarger/ on LinkedIn Lou Schmidt @loudoggeek on X https://www.linkedin.com/in/louis-schmidt-b102446/ on LinkedIn ---------------------------------------- Hosted on Acast. See acast.com/privacy [https://acast.com/privacy] for more information.

Microsoft Exchange Zero-Day: No Patch, Active Exploitation, Major Risk

A newly disclosed Microsoft Exchange vulnerability is actively being exploited in the wild, and there’s still no permanent patch available. In this episode of IT SPARC Cast – CVE of the Week, John and Lou break down CVE-2026-42897, explain how attackers can exploit Outlook Web Access through malicious emails, and discuss why temporary mitigations may not be enough for organizations still running on-prem Exchange. ⸻ 📄 Show Notes 🚨 CVE of the Week: Microsoft Exchange / Outlook Web Access Exploit This week’s episode focuses on CVE-2026-42897, a high-severity vulnerability affecting: * Microsoft Exchange Server 2016 * Microsoft Exchange Server 2019 * Exchange Subscription Edition The vulnerability is a cross-site scripting (XSS) and spoofing flaw impacting Outlook Web Access (OWA). ⸻ ⚠️ How the Attack Works Attackers send specially crafted emails that execute malicious JavaScript when opened through Outlook Web Access. Potential impacts include: * Session hijacking * Browser-based code execution * Exchange session theft * Spoofing attacks The vulnerability is already being actively exploited in the wild. ⸻ 🌐 Who Is Affected? This impacts on-prem Exchange deployments only. Cloud-hosted Exchange Online environments are not currently believed to be affected. Organizations most at risk include: * Enterprises with legacy Exchange infrastructure * Organizations avoiding cloud email hosting * Remote-access-heavy environments relying on OWA ⸻ 🛠️ Mitigation Steps for CVE-2026-42897 ✅ 1️⃣ Apply Microsoft Emergency Mitigations Microsoft has released temporary protections through: * Exchange Emergency Mitigation Service (EEMS) * URL rewrite mitigation rules Apply these immediately. ⚠️ Important: These mitigations are pattern-based and may not block future modified exploits. ⸻ ✅ 2️⃣ Consider Disabling Outlook Web Access (OWA) If operationally possible: * Disable OWA temporarily * Require users to use the Outlook desktop client instead This significantly reduces exposure. ⸻ ✅ 3️⃣ Prepare for Operational Side Effects Known mitigation side effects include: * Calendar printing failures * Inline image rendering problems * Increased help desk tickets Organizations should proactively communicate these issues to users. ⸻ ✅ 4️⃣ Patch Immediately When Available At recording time: * No permanent patch exists yet * Apply the official patch immediately once released This is not a vulnerability where delayed patching is safe. ⸻ 🔒 Security Takeaways This vulnerability reinforces several growing cybersecurity realities: * On-prem infrastructure carries operational security burdens * Browser-based attacks remain highly effective * Temporary mitigations are not substitutes for permanent fixes John and Lou also discuss how attackers increasingly chain vulnerabilities together and how AI-assisted exploit development is accelerating the speed of attacks. ⸻ 💬 Listener Feedback Thanks to listener “ZZZZ” on YouTube for pushing back on last week’s discussion around passwords stored in clear text memory. The discussion highlights an important point: * Many vulnerabilities are low risk for average users * But become extremely dangerous for high-value targets such as executives and organizations with sensitive data ⸻ 📣 Wrap Up Are organizations moving away from on-prem Exchange fast enough, or are these vulnerabilities making the case for cloud migration even stronger? 📧 feedback@itsparccast.com 🐦 @itsparccast on X ⸻ 🔗 Social Links IT SPARC Cast @ITSPARCCast on X https://www.linkedin.com/company/sparc-sales/ on LinkedIn John Barger @john_Video on X https://www.linkedin.com/in/johnbarger/ on LinkedIn Lou Schmidt @loudoggeek on X https://www.linkedin.com/in/louis-schmidt-b102446/ on LinkedIn ---------------------------------------- Hosted on Acast. See acast.com/privacy [https://acast.com/privacy] for more information.

OpenAI’s Daybreak, Google’s AI Laptop Push, and Cisco’s AI Fingerprinting Tool

In this episode of IT SPARC Cast - News Bytes, John & Lou break down the growing intersection of AI, cybersecurity, and enterprise infrastructure. OpenAI enters the AI security space with Daybreak, Google unveils a new AI-native laptop platform called Googlebook, and Cisco releases an open source tool designed to trace the origins of AI models. The discussion focuses on how AI is rapidly moving from experimentation into operational reality. From AI-assisted security operations to AI-centric hardware and supply chain validation for large language models, this episode explores the practical implications these technologies will have on enterprise IT teams over the next few years. ⸻ 📌 Show Notes 00:00 – Intro This week’s episode covers AI-powered cybersecurity, Google’s next-generation laptop strategy, and growing concerns around AI model provenance and trust. ⸻ 📰 News Bytes 00:44 – OpenAI Launches Daybreak OpenAI launched Daybreak, an AI-powered vulnerability detection and patch validation platform designed to help overwhelmed security teams handle rising alert volumes and faster-moving threats. The system uses AI agents to analyze alerts, correlate activity, assist with incident response, and reduce analyst fatigue. John & Lou discuss how AI works best as a force multiplier for security teams—not as a replacement for experienced analysts. Key takeaways: * AI excels at repetitive security analysis tasks * Human oversight is still critical * Over-automation increases operational risk https://thehackernews.com/2026/05/openai-launches-daybreak-for-ai-powered.html [https://thehackernews.com/2026/05/openai-launches-daybreak-for-ai-powered.html] ⸻ 06:39 – Google Unveils Googlebook Google announced “Googlebook,” a new category of AI-native laptops deeply integrated with Gemini AI and built on a combined Android/Chrome OS platform. The devices aim to compete directly with AI-focused Windows PCs and MacBooks while emphasizing web-first workflows, Android integration, and AI-enhanced interfaces like the new “Magic Pointer.” Key considerations: * Enterprise apps are increasingly web-based * OS dependency continues to decline * AI-native devices may reshape endpoint strategy https://techcrunch.com/2026/05/12/google-unveils-googlebooks-a-new-line-of-ai-native-laptops/ ⸻ 13:04 – Cisco Releases Open Source AI Provenance Tool Cisco released an open source tool designed to determine the origins and lineage of AI models. The tool can compare models directly or scan against known fingerprints to identify derivative training sources. The goal is improving AI supply chain security by detecting repackaged models, inherited vulnerabilities, licensing issues, and potentially poisoned AI systems. Key implications: * AI supply chain security is becoming critical * Organizations need visibility into model origins * Provenance tracking may become standard practice https://github.com/cisco-ai-defense/model-provenance-kit https://blogs.cisco.com/ai/model-provenance-kit ⸻ 📬 17:43 – Mail Bag Listener feedback revisits Microsoft Edge storing passwords in plaintext memory and sparks a broader discussion around practical enterprise security decisions, browser trust, and balancing usability against risk. ⸻ 🔚 19:35 – Wrap Up As AI rapidly expands into security, infrastructure, and endpoint computing, organizations must balance innovation with governance and operational discipline. The future of enterprise IT will depend not just on adopting AI—but understanding and securing it properly. ⸻ 🌐 Social Links IT SPARC Cast @ITSPARCCast on X https://www.linkedin.com/company/sparc-sales/ on LinkedIn John Barger @john_Video on X https://www.linkedin.com/in/johnbarger/ on LinkedIn Lou Schmidt @loudoggeek on X https://www.linkedin.com/in/louis-schmidt-b102446/ on LinkedIn ---------------------------------------- Hosted on Acast. See acast.com/privacy [https://acast.com/privacy] for more information.