Cloud and Cybersecurity News

☁️ Cloud   Organizations are transitioning toward decentralized multi-cloud observability by shifting from traditional diagnostic collection to continuous, high-fidelity platform metric streaming using Data Collection Rules.   🔒 Cybersecurity   Adversaries are actively weaponizing software vulnerabilities within hours using automated artificial intelligence, shifting the enterprise perimeter focus from endpoint defense toward continuous identity authentication and API logging.   🤖 AI/ML   The Model Context Protocol (MCP) has emerged as an open enterprise standard that eliminates custom data-access layers by allowing autonomous AI agents to natively query, discover, and reason about relational database schemas.   Source:  https://www.linkedin.com/pulse/weekly-news-update-1262026-eyal-estrin--qzq6f/   #Cloud #Cybersecurity #AI

☁️ Cloud Microsoft outlines a GA architectural blueprint for Azure Cloud Landing Zones that implements a multi-hub topology to segregate North-South internet ingress from internal East-West movements. 🔒 Cybersecurity Anthropic's threat intelligence analysis reveals a technical shift from initial entry tactics toward post-compromise lateral movement and advanced account discovery. 🤖 AI/ML Amazon Bedrock announces the GA release of OpenAI frontier reasoning and Codex models to handle complex agentic workflows within secure enterprise environments.   Source: https://www.linkedin.com/pulse/weekly-news-update-562026-eyal-estrin--froef/ #Cloud #Cybersecurity #AI

☁️ Cloud AWS launched Random Node Grouping data center networks using flat topologies based on random graph theory to cut electricity consumption by 40%. 🔒 Cybersecurity The Tamnoon State of Cloud Remediation 2026 report revealed a widening operational gap that leaves 53% of all cloud security alerts open and unresolved. 🤖 AI/ML Anthropic released Claude Opus 4.8 under General Availability, featuring a fourfold reduction in code-generation errors and enhanced long-horizon agentic workflows. Source: https://www.linkedin.com/pulse/weekly-news-update-2952026-eyal-estrin--3dquf/ #Cloud #Cybersecurity #AI

🌐 Cloud: Google GKE Agent Sandbox Goes Live Google launched its Agent Sandbox on GKE into General Availability, tackling the security risks of autonomous AI agents. By utilizing gVisor kernel isolation and default-deny network policies, it safely confines untrusted, LLM-generated code. Optimized for speed, it provisions up to 300 sandboxes per second with sub-200-millisecond latency.   🔒 Cybersecurity: Vulnerability Exploits Overtake Credentials The 2026 Verizon Data Breach Investigations Report revealed a historic shift: for the first time in 19 years, software vulnerability exploitation is the leading entry point for breaches (31%). Attackers are aggressively using AI to weaponize known flaws within hours, while corporate patch times lag significantly behind with a median of 43 days.   🤖 AI/ML: Google Drops Gemini 3.5 Flash Google released Gemini 3.5 Flash directly to production, delivering flagship-tier reasoning at four times the speed of competing frontier models. The architecture is explicitly engineered for dense, long-horizon workflows and collaborative subagent loops, pairing natively with the Antigravity platform to orchestrate concurrent AI agents. https://www.linkedin.com/pulse/weekly-news-update-2252026-eyal-estrin--lzinf/ #Cloud #Cybersecurity #AI

🌐 Cloud: The Offloading and Containerization Era Microsoft's General Availability of the Next Generation of Azure Boost marks a major leap in performance. By offloading foundational virtualization, networking, and storage processes from the host CPU onto purpose-built hardware, it achieves up to 800k IOPS and 12.5 GB/s in storage throughput. This hardware isolation drastically lowers latency for data-heavy workloads while shrinking the host attack surface. Concurrently, tools like AWS Transform are matching this push by letting enterprises completely skip simple "lift-and-shift" migrations, automatically replatforming legacy workloads straight into secure, production-ready container architectures. 🔒 Cybersecurity: The Expansion of "Agentic" and Identity Risk The State of AI Agent Security 2026 details a massive shift from simple data leakage chatbot risks to Agentic Risk. Attackers are now targeting the internal logic of autonomous agents using chain-of-thought prompt injection to force unauthorized administrative actions. This risk is amplified by a massive Identity Visibility Gap highlighted by Zoho, where 74% of organizations admit they cannot fully account for workforce identities or orphaned accounts, leaving automated business workflows to operate dangerously outside traditional Identity and Access Management (IAM) frameworks. 🤖 AI/ML: The Rise of Autonomous Control and Principled Alignment Microsoft Copilot Studio's General Availability of computer-using agents allows AI to navigate legacy software interfaces without APIs by mimicking human clicks, typing, and scrolling. As systems gain this physical agency, safety frameworks are scrambling to keep up. Anthropic's "Principled Alignment Training" research addresses this exact friction point. Their findings show that teaching models why an action is ethically right or wrong via step-by-step reasoning is significantly more effective at stopping autonomous tool misuse than simply training them on examples of good behavior. https://www.linkedin.com/pulse/weekly-news-update-1552026-eyal-estrin--mvvcf/

Cloud: The "Data Plane" Audit Revolution AWS EventBridge has officially launched General Availability (GA) for Data Plane logging to CloudTrail. For the first time, organizations can move beyond basic management logging to record high-volume PutEvents activities. This is a massive win for governance, providing a verifiable audit trail to detect unauthorized event injections and solve the "black box" problem in complex, event-driven architectures. Cybersecurity: Surviving the "AI Vulnerability Storm" As AI collapses the time between vulnerability discovery and weaponization, the Cloud Security Alliance (CSA) has released its new framework for building "Mythos-ready" security programs. The focus is shifting from traditional patching to VulnOps and autonomous security agents. The key takeaway? Attacker asymmetry is real, and defenders must operationalize AI-driven transaction flow discovery and identity-centric Software-Defined Perimeters (SDP) to stay ahead of machine-speed threats.AI/ML: The Rise of the MCP StandardInfrastructure is finally learning to "talk" to AI. Both AWS and Azure have announced the General Availability of the Model Context Protocol (MCP) Server. This standardized protocol allows AI agents to natively discover, query, and reason about cloud resources (like S3, Lambda, and ARM) without custom glue code. It transforms the cloud from a collection of APIs into a "context-aware" playground for autonomous agents, governed by existing enterprise security boundaries. https://www.linkedin.com/pulse/weekly-news-update-852026-eyal-estrin--tatnf/

Cloud: From Infrastructure to Managed Intelligence The Unification of AI and Infrastructure: The Wiz State of AI in the Cloud 2026 report confirms that AI has moved from experimental tooling to a default component of cloud architecture, though this transition has introduced significant "Shadow AI" risks—with 81% of environments now using managed AI services while 25% of organizations lack visibility into these active integrations. Cybersecurity: The Identity and Skills Tipping PointOperationalizing the Identity Attack Surface: Identity Attack Path Management (APM) has reached a critical tipping point where 75% of enterprises are increasing spending to move beyond mere visibility toward the complex task of remediating non-human identity risks—specifically the explosion of AI agents, now used by 88% of organizations, which create exploited delegated permissions. AI/ML: Breaking Model Lock-in and Scaling Planetary AnalysisOpen Access and Multi-Cloud Sovereignty: The landscape is shifting toward extreme flexibility, highlighted by the General Availability of Gemma 4 on AWS SageMaker and Microsoft's new "Bring Your Own Model" (BYOM) capability in Foundry Agent Service; meanwhile, Google has democratized planetary-scale analysis by integrating Google Earth AI models and datasets directly into BigQuery.https://www.linkedin.com/pulse/weekly-news-update-152026-eyal-estrin--66dif/

Cloud: The Shift to Resilient, Scalable Data Fabrics Google has launched the Virgo network fabric, a "megascale" architecture designed specifically for the massive communication demands of modern AI. By utilizing Optical Circuit Switching (OCS), Virgo allows data centers to dynamically reconfigure network paths at the fiber level. This reduces power consumption and improves reliability by automatically bypassing faulty hardware. For architects, this represents a shift toward a unified AI interconnect where TPUs, GPUs, and storage share resources without the traditional bottlenecks of tiered networks. Cybersecurity: Identity as the New Perimeter The Cisco Talos 2025 Year in Review confirms that attackers have moved away from malware-heavy entries in favor of identity-centric tactics. Credential theft and session hijacking are now the primary drivers of breaches, effectively turning "living off the cloud" into a standard operating procedure for threat actors. To counter this, security teams must pivot from traditional endpoint defense to monitoring API activities and cloud-native service permissions, as automated ransomware now exfiltrates data faster than manual investigation can detect. AI/ML: The Rise of Autonomous Agentic Workflows The release of GPT-5.5 marks a transition from simple chatbots to autonomous agents capable of independent multi-step planning and self-correction. This "agentic" shift is mirrored in the infrastructure layer, with tools like Gemini Embedding 2 providing the semantic foundation for Retrieval-Augmented Generation (RAG). However, this autonomy brings risk; the Cloud Security Alliance reports that 65% of enterprises experienced an AI agent-related incident last year, with shadow AI agents frequently appearing in internal automation environments. https://www.linkedin.com/pulse/weekly-news-update-2442026-eyal-estrin--cb7df/

Cloud: The Arrival of Native Multicloud Networking The headline in cloud infrastructure is the general availability of AWS Interconnect. This service marks a strategic shift by AWS to simplify "last-mile" connectivity to other providers. By leveraging the AWS global backbone, architects can now establish high-speed, direct links to rival clouds without managing complex third-party circuits or brittle VPN tunnels. This move effectively treats multicloud integration as a native networking feature rather than a logistical hurdle. Cybersecurity: The Race Against AI-Powered Exploits The central theme in security is the "Patch Sound Barrier." Research highlights that AI is now accelerating vulnerability discovery and exploit chaining to a speed that traditional human-led remediation cannot match. To counter this, industry leaders are advocating for a defensive shift toward autonomous "self-red-teaming." The strategy involves deploying frontier models to scan codebases and generate patches within 24-hour windows, essentially using the same AI tools as attackers to harden the perimeter before exploits can be realized. AI/ML: The Rise of Agentic Engineering The most significant trend in AI is the transition from simple chat interfaces to autonomous agentic workflows. With the launch of Claude Opus 4.7 and its "Advisor Strategy," models can now self-verify code and orchestrate multi-step engineering tasks with minimal human intervention. This is supported by new specialized tooling like GPT-5.4-Cyber and domain-specific models like GPT-Rosalind for life sciences. These developments signal a move toward "Expert AI" that doesn't just suggest content but actively executes complex research and development cycles. https://www.linkedin.com/pulse/weekly-news-update-1742026-eyal-estrin--peu8f/

Cloud: Amazon S3 Becomes a File System The long-standing architectural wall between S3 buckets and local file systems has collapsed. With the launch of Amazon S3 Files, AWS now provides a kernel-level, high-performance way to mount S3 buckets directly onto Linux and Windows. Unlike older workarounds that suffered from "eventual consistency" lag, this native feature brings strong consistency and local caching. For architects, this means legacy applications can finally treat the cloud's largest storage pool as a local disk without rewriting a single line of code. Cybersecurity: The "Agentic" Visibility Crisis As AI agents move from chat interfaces to the "Action Layer" of the enterprise—making API calls and executing business logic—security is falling behind. The 1H 2026 State of AI and API Security Report highlights a massive blind spot: nearly half of organizations cannot monitor machine-to-machine traffic. With traditional tools like WAFs proving ineffective against authenticated API abuse, the industry is entering a "visibility crisis" where autonomous agents are operating in trusted systems with almost zero oversight. AI/ML: Anthropic's Mythos and the Future of Defense The release of Claude Mythos via Project Glasswing marks a turning point where AI is no longer just a coding assistant but a proactive security researcher. Integrated into the Linux Foundation's infrastructure, Mythos is designed for autonomous vulnerability discovery and remediation in open-source codebases. While it can solve 44% of complex Capture the Flag (CTF) challenges, its real value lies in its ability to map attack surfaces and generate functional patches for zero-day flaws before they can be exploited. https://www.linkedin.com/pulse/weekly-news-update-1042026-eyal-estrin--pxxjf/

Cloud: The AWS DevOps Agent Goes GA The AWS DevOps Agent is now generally available, marking a shift toward unified hybrid cloud management. This single agent streamlines operations by automating software deployment, patching, and configuration across both Amazon EC2 and on-premises servers. Crucially, it supports secure, private VPC connectivity and allows for the integration of agentic AI to handle autonomous incident response and routine maintenance. Cybersecurity: The Axios Supply Chain Compromise A sophisticated supply chain attack has targeted the widely used axios NPM package. Attributed to the North Korean threat actor UNC1069, the breach involved hijacking a maintainer's account to embed a cross-platform remote access trojan (RAT) called WAVESHAPER.V2. With over 100 million weekly downloads, this incident underscores the urgent need for teams to audit build logs and pin dependencies to verified safe versions. AI/ML: The Launch of Gemma 4 Google has released Gemma 4, a new family of open models built on Gemini 3 technology. These models are engineered for "intelligence-per-parameter," featuring native multimodal capabilities and advanced agentic skills like function calling and structured JSON output. With variants ranging from a 26B Mixture of Experts (MoE) to edge-optimized "Effective" models, it brings frontier-level reasoning directly to mobile and IoT devices. https://www.linkedin.com/pulse/weekly-news-update-342026-eyal-estrin--btscf/

Cloud: Standardizing AI OrchestrationThe Cloud Native Computing Foundation (CNCF) has accepted llm-d into its Sandbox, providing a Kubernetes-native framework for high-performance distributed LLM inference. Supported by Google Cloud, IBM, and NVIDIA, this project allows architects to automate scaling and fault tolerance for AI workloads across vendor-neutral infrastructure.Cybersecurity: Hardening Model InferenceAs Large Language Models move into production, security is shifting toward resource-level isolation within the orchestration layer. By leveraging Kubernetes-native scheduling for model deployments, organizations can now enforce consistent security boundaries and identity-based access controls directly on GPU-accelerated clusters.AI/ML: The Efficiency of Disaggregated PrefillA major technical milestone in the latest inference frameworks is the implementation of disaggregated prefill and decode. This specific optimization separates the initial prompt processing from the token generation phase, significantly reducing "Time to First Token" (TTFT) and maximizing GPU utilization for enterprise-scale reasoning tasks. https://creators.spotify.com/pod/show/1ver9PijJI9yUM83Bvnjov/episode/5DLSJFGsguaX6mIT44OU0v/wizard

Cloud: The Evolution of Foundations Amazon S3 marks its 20th anniversary by transitioning from simple object storage into a foundational ecosystem for data and AI. Key updates reaching General Availability (GA) include S3 Tables for managed Apache Iceberg optimization and S3 Vectors, which provides native storage for semantic search and Retrieval-Augmented Generation (RAG) with sub-100ms latency. Cybersecurity: The Reality of Secrets Sprawl The GitGuardian State of Secrets Sprawl 2026 report highlights a record 29 million leaked credentials on public GitHub. A major driver is the ungoverned use of AI assistants; leaked secrets for AI services surged 81% year-over-year, and internal repositories are now six times more likely to contain hardcoded secrets than public ones. AI/ML: Agentic Scaling and Infrastructure Innovation is shifting toward production-grade "agentic" workflows. OpenAI has released GPT-5.4 mini and nano, optimized for high-volume tasks and low-latency computer-use. Meanwhile, Anthropic has moved its 1-million-token context window to General Availability for Claude 4.6 (Opus and Sonnet), enabling high-fidelity reasoning across massive datasets and entire codebases. https://www.linkedin.com/pulse/weekly-news-update-2132026-eyal-estrin--y4oqf/

CloudAWS is transforming Security Hub into a unified multicloud operations platform. Moving beyond simple posture management, the service now integrates third-party signals from partners like CrowdStrike and Okta into a single data layer. This allows teams to manage risk and billing across AWS and external workloads without switching consoles.CybersecurityThe 2026 State of Passwordless Identity Assurance report reveals a major shift: AI-driven automation has officially surpassed stolen credentials as the primary identity threat. With 87% of organizations facing deepfake attacks, the "Velocity Paradox" means AI can now exfiltrate data faster than human teams can respond, making traditional point-in-time authentication obsolete.AI/MLMicrosoft has launched Agent 365, a centralized control plane designed to govern autonomous AI "coworkers." By assigning every agent a unique Entra Agent ID, the system treats AI as a privileged machine identity. This ensures that autonomous agents follow Zero Trust policies and prevents "shadow AI" from accessing sensitive corporate data without oversight.https://www.linkedin.com/pulse/weekly-news-update-1332026-eyal-estrin--fx1rf/

Cloud: Deep dive into this week's biggest cloud moves, from new managed services and resiliency patterns to what they mean for building secure, scalable architectures in AWS, Azure, and GCP.Cybersecurity: Breakdown of the latest threats and regulations shaping cyber risk in 2026, including evolving ransomware campaigns, supply chain attacks, and how security teams can strengthen cloud-native defenses and governance.AI/ML: Practical look at how modern AI models and agents are changing security operations, from AI-assisted threat detection and incident response to the new challenges of prompt governance and model safety.https://www.linkedin.com/pulse/weekly-news-update-632026-eyal-estrin--u24rf/

Cloud InfrastructureThe focus has shifted heavily toward sovereign cloud solutions. Major providers are rolling out localized infrastructure to meet strict data residency requirements, moving away from a one-size-fits-all global model. CybersecurityIdentity is the new perimeter. Organizations are moving beyond simple MFA to embrace continuous authentication, reacting to a surge in sophisticated session-hijacking attacks that bypass traditional login protections.AI & Machine LearningThe industry is moving from massive, general-purpose models toward "Small Language Models" (SLMs). These efficient, task-specific tools allow companies to run powerful AI locally or on the edge without the massive compute costs of larger counterparts.https://www.linkedin.com/pulse/weekly-news-update-2722026-eyal-estrin--yasnf/

Cloud: The npm Supply Chain Breach A significant breach in the npm registry has exposed sensitive cloud secrets, highlighting a persistent weakness in how we manage dependencies. For Cloud Architects and CISOs, this serves as a blunt reminder that your security is only as strong as your third-party code. The takeaway: it is time to move beyond trust and implement rigorous audits of all external dependencies and service connections. Cybersecurity: The Rise of Autonomous AI Threats We are entering a new era where autonomous AI threats are no longer theoretical. Global data leaks involving billions of records have surged as companies struggle to secure privileged user accounts against AI-driven intrusions. As attackers move from manual exploits to automated, self-evolving scripts, traditional defense perimeters are proving insufficient. AI/ML: The Asymmetry of Risk AI has officially become a force multiplier for attackers, compressing weeks of reconnaissance into mere days. The current challenge lies in the asymmetry of defense: while attackers can iterate rapidly with little oversight, defensive AI must operate within strict governance and human review frameworks. To close this gap, organizations must prioritize real-time, machine-readable visibility of their entire attack surface. https://www.linkedin.com/pulse/weekly-news-update-2022026-eyal-estrin--ajsuf/

Cloud: The Shift to "Clean Sky" Architecture Cloud providers are moving beyond simple infrastructure toward sovereign cloud frameworks. Major players like AWS and Azure have introduced deeper automation for regional data residency, ensuring that metadata—not just user data—remains within specific geographic borders to meet tightening EU and global regulations. Cybersecurity: The IPv6 Blind Spot As organizations transition away from IPv4, a massive security gap is emerging in IPv6 visibility. Many traditional monitoring tools fail to map the expanded address space, allowing attackers to bypass perimeter defenses. The takeaway is clear: if you aren't scanning your IPv6 assets with the same rigors as your legacy network, you have an invisible "back door" open. AI/ML: Speeding Up the Kill Chain Artificial Intelligence has officially transitioned from a "possibility" to a force multiplier for attackers. New data shows that AI is being used to compress the cyber-attack lifecycle—turning what used to be weeks of reconnaissance and lateral movement into a matter of days. Defenders are now forced to adopt machine-speed response times just to stay level with automated exploits. https://www.linkedin.com/pulse/weekly-news-update-1322026-eyal-estrin--u7cof/

Cloud: Infrastructure & Transparency The shift toward Hybrid Computing is accelerating as organizations realize that running continuous AI workloads solely in the cloud is becoming cost prohibitive. We also see a growing tension between cloud providers and local communities; for example, Microsoft's recent data center expansion challenges highlight that cloud growth now requires greater transparency and community buy-in regarding land and resource use. Cybersecurity: The High Cost of Noise Security teams are facing a "Vulnerability Noise Tax," where investigating false positives costs an average of $1,000 and 12 hours per alert. This inefficiency is compounded by a massive security advisory for SolarWinds Web Help Desk, which includes four critical Remote Code Execution (RCE) flaws. Immediate updates to version 2026.1 are essential to prevent attackers from bypassing authentication entirely. AI/ML: Securing Autonomous Agents The industry is moving from traditional LLMs to Agentic AI—autonomous bots that act as independent applications. This transition creates a new "identity chain" problem: how do we properly scope and manage permissions for a bot that makes real-time decisions via APIs? Implementing a robust AI Governance Framework is now a prerequisite for any enterprise moving these agents into production. https://www.linkedin.com/pulse/weekly-news-update-622026-eyal-estrin--xc5lf/

https://www.linkedin.com/pulse/weekly-news-update-3012026-eyal-estrin--xwmzf/