The April 2026 Windows update cycle includes several changes that matter to IT administrators.

Read More

OpenAI, together with AMD, Broadcom, Microsoft, and NVIDIA, published a new paper describing MRC (Multipath Reliable Connection), a new networking protocol designed for large AI training clusters. MRC addresses two of the most critical problems in these networks: traffic congestion and link failures. The protocol is already deployed in production at OpenAI and Microsoft data centers. The specification is freely available through the Open Compute Project (OCP) under an open license.

Read More

Microsoft 365 Copilot Cowork moves beyond chat-based AI responses by acting as an autonomous agent—an AI that can perform multi-step tasks on your behalf across Outlook, Teams, Word, Excel, and other Microsoft 365 services. Cowork is built on top of Work IQ, Microsoft’s intelligence layer that reads your emails, meetings, files, and organizational data to provide context for actions. As of May 2026, Cowork is available through the Frontier preview program and has expanded to include mobile support, custom skills, plugins, and integration with Agent 365. Access requires a Microsoft 365 Copilot license and enrollment in the Frontier early-access program.

Read More

Microsoft announced three updates for its cloud desktop products in May 2026: a 20% price cut for Windows 365 Business, a rename of Windows 365 Frontline to Windows 365 Flex, and a public preview of Azure Virtual Desktop (AVD) Hybrid, which lets you run cloud-managed virtual desktops on your own on-premises servers. This article focuses on the technical aspects of these changes, particularly AVD Hybrid, which introduces meaningful new infrastructure options for organizations that cannot fully migrate workloads to Azure.

Read More

Microsoft 365 Backup now lets you recover individual files and folders from SharePoint and OneDrive backups without rolling back an entire site or account. This granular restore feature became generally available in late April 2026. You need the SharePoint Backup Administrator role to use it, and the service charges $0.15 per GB per month for protected data. This article explains how the feature works, its limitations, and how to perform a restore.

Read More

Former OpenAI Chief Technology Officer Mira Murati testified under oath that CEO Sam Altman provided false information regarding the safety protocols for a new AI model. During a video deposition for the Musk v. Altman trial, Murati stated that Altman incorrectly claimed the legal department had exempted a specific model from review by the company’s deployment safety board. She discovered the discrepancy after consulting with General Counsel Jason Kwon and subsequently ensured the model underwent the n…Read more

Microsoft is introducing Point-in-Time Restore to Windows 11, a modern recovery feature designed to provide more granular system snapshots than the legacy System Restore tool. Unlike its predecessor, which primarily focuses on system files and registry settings, this new utility captures user files, application data, passwords, and security certificates. The feature utilizes the Volume Shadow Copy Service (VSS) to create these comprehensive snapshots without interrupting active workflows.

The tool operates…Read more

Cisco has released security updates to address a high-severity denial-of-service vulnerability affecting its Crosswork Network Controller and Network Services Orchestrator platforms. The flaw, tracked as CVE-2026-20188, allows unauthenticated remote attackers to exhaust system connection resources through low-complexity attacks. Impacted systems become completely unresponsive to legitimate users and dependent services, necessitating a physical or manual reboot to restore functionality.

The vulnerability stems from…Read more

A routine key rollover for Germany’s .de top-level domain (TLD) on May 5, 2026, resulted in the publication of invalid DNSSEC signatures. DNSSEC provides cryptographic integrity to DNS records, and its strict validation requirements forced resolvers worldwide to reject .de queries as “bogus” to prevent potential tampering. This misconfiguration threatened the reachability of millions of domains, causing a massive spike in SERVFAIL errors as cached records expired across global recursive resolvers.

To m…Read more

Microsoft is accelerating the transition to Windows on Arm by collaborating with major hardware vendors to ensure robust printing capabilities. This initiative centers on the Modern Print Strategy, which utilizes the Internet Printing Protocol (IPP) and the Windows Protected Print mode to eliminate the need for legacy third-party drivers. By moving toward a driverless architecture, administrators can reduce security risks and simplify deployment across diverse hardware environments.

Major industry players including…Read more

Microsoft has expanded Copilot Cowork, an agentic AI layer designed to execute multi-step workflows autonomously across the Microsoft 365 ecosystem. Unlike standard conversational AI, this system uses a specialized intelligence layer called Work IQ to plan and perform background tasks like document generation and meeting coordination. The service is currently available through the Frontier early-access program and is built in collaboration with Anthropic to leverage advanced reasoning capabilities.

The latest…Read more

OpenAI has introduced the ChatGPT Futures Class of 2026, a program designed to support the first generation of students who have completed their entire college education alongside generative AI. The initiative recognizes 26 student innovators from over 20 global institutions, including Georgia Tech, Vanderbilt, and Oxford, who are using AI to solve complex real-world problems. These honorees were selected for their “high-agency” approach, demonstrating how individuals can now prototype ideas and build impactful…Read more

Anthropic has secured a major partnership with SpaceX to utilize the full capacity of the Colossus 1 data center in Memphis, Tennessee. This agreement provides immediate access to over 300 megawatts of power and more than 220,000 NVIDIA GPUs, including H100, H200, and next-generation Blackwell GB200 accelerators. The influx of hardware allows the company to significantly scale its infrastructure to meet the growing demand for its Claude artificial intelligence models.

Effective immediately, Anthropic has…Read more

Anthropic has expanded its Claude Managed Agents platform with new features designed to automate complex workflows and enable autonomous self-improvement. Managed Agents is a suite of APIs that provides the infrastructure for running AI agents, including secure sandboxing, long-running sessions, and persistent file systems. The latest update introduces “dreaming,” a research preview feature that allows agents to review past session transcripts and memory stores to identify patterns and refine their…Read more

Microsoft is launching its first voluntary retirement program in its 51-year history, targeting approximately 8,750 long-serving U.S. employees. The initiative uses a “Rule of 70” formula, where eligibility is granted to staff whose combined age and years of service equal at least 70. This program specifically excludes employees on sales incentive plans and those in senior leadership roles above the senior director level.

The buyout package includes a lump sum cash payment based on tenure, five years of…Read more

DeepSeek is reportedly in discussions for its first external funding round, which could value the Chinese artificial intelligence lab at approximately $45 billion. This figure represents a massive surge from a $20 billion valuation discussed only weeks prior and a $10 billion target initially set in mid-April. The rapid escalation reflects intense interest in the firm, which gained global recognition for developing high-performance large language models at a fraction of the cost required by Western…Read more

Microsoft Edge has been found to decrypt every saved credential at startup and maintain them in plaintext within the browser’s process memory. This behavior occurs regardless of whether the user visits the websites associated with those credentials during their session. While Edge requires re-authentication to view passwords through the settings interface, the underlying process memory remains populated with the sensitive data in an unencrypted state.

Security researchers note that this implementation…Read more

The server integrates with existing AWS governance tools like Identity and Access…Read more

The toolkit features a…Read more

Google has introduced multi-token prediction (MTP) drafters for its Gemma 4 open model family to address memory-bandwidth bottlenecks during text generation. Standard large language models (LLMs) typically generate text one token at a time, which leaves the processor idle while waiting for massive parameter files to load from memory. This new technology utilizes that idle capacity by running a lightweight auxiliary model alongside the primary model to suggest multiple future tokens…Read more

Atlassian is expanding access to its Teamwork Graph, a data layer that maps relationships between people, projects, and goals across its software ecosystem. This foundation provides the essential context required for AI agents to understand organizational structures and dependencies beyond simple keyword matching. By moving away from a closed system, the company aims to position this graph as a universal “system of work” for modern enterprises.

The company has introduced a new Command Line Interface (CLI) and a…Read more

Hewlett Packard Enterprise has transitioned its self-driving network vision into production by deploying autonomous agents across the HPE Mist and Aruba Central platforms. These agentic AI capabilities move beyond simple alerts to perform closed-loop remediation of high-frequency issues like wireless congestion, VLAN misconfigurations, and rogue DHCP servers. By utilizing a microservices architecture and a Large Experience Model, the system can predict performance degradation in collaboration…Read more

Microsoft has introduced a native NVMe storage driver, nvmedisk.sys, in Windows Server 2025 to replace the legacy SCSI translation layer. This older architecture forced modern SSDs to emulate Small Computer System Interface (SCSI) commands, a protocol originally designed for mechanical hard drives. By removing this translation step, the new driver reduces CPU overhead and significantly improves Input/Output Operations Per Second (IOPS) for high-performance storage.

Although officially documented only…Read more

Adobe has introduced a new artificial intelligence productivity agent for Acrobat designed to automate the creation, summarization, and sharing of digital documents. This agentic tool orchestrates multiple AI models to transform static PDFs into dynamic content, such as presentations, podcasts, and social media posts, through natural language prompts. The technology is primarily integrated into two new offerings: Acrobat Express, a lightweight content generation tool, and Acrobat Studio, which combines advanced PDF…Read more

Apple is reportedly preparing a significant shift in its artificial intelligence strategy by allowing users to replace or supplement Siri with third-party models. This new framework, internally referred to as “Extensions,” is expected to debut with the release of iOS 27, iPadOS 27, and macOS 27 later this year. The system will enable generative AI services from installed App Store applications to power core features like Writing Tools, Image Playground, and voice assistant responses.

The move marks a departure from…Read more

Google has updated its AI Overviews and AI Mode features to integrate firsthand perspectives and improve source transparency within search results. A new “Expert Advice” section now displays brief remarks and discussion snippets from social media and online forums, complete with creator names and community handles. These additions aim to provide users with authentic human experiences alongside AI-generated summaries for complex or subjective queries.

The update also introduces a dedicated feature that…Read more

OpenAI has launched B2B Signals to track how organizations transition from basic AI experimentation to deep, production-level integration. Research indicates that “frontier firms” now utilize 3.5 times more intelligence per worker than typical companies, a gap that has widened significantly over the past year. This advantage is driven primarily by the depth and complexity of interactions rather than simple message volume, as leading firms move beyond basic chat queries toward executing multi-step…Read more

The Venomous#Helper phishing campaign has targeted over 80 organizations since April 2025, primarily focusing on the United States and Western Europe. Attackers utilize emails impersonating the U.S. Social Security Administration to trick victims into downloading signed, malicious executables from compromised legitimate domains. These files install a dual-channel architecture using legitimate Remote Monitoring and Management (RMM) tools, specifically SimpleHelp and ScreenConnect, to ensure…Read more

Hewlett Packard Enterprise has transitioned its networking strategy from passive observation to active autonomy by integrating agentic AI across the HPE Aruba Networking Central and HPE Juniper Networking Mist platforms. This shift introduces “self-driving actions” that allow the system to move beyond merely flagging issues to executing fixes without manual intervention. The architecture utilizes specialized AI agents and a microservices foundation to handle high-frequency problems like wireless congestion, VLAN…Read more

AMD is transitioning its EPYC server processor strategy from a general-purpose approach to a highly segmented portfolio designed for specific data center roles. The upcoming Zen 6-based “Venice” family, expected in 2026, will introduce specialized SKUs such as “Verona” to specifically address AI infrastructure and agentic AI tasks. This shift aims to compete with custom silicon programs by offering tailored configurations for throughput, power efficiency, and low-latency orchestration.

The Zen 6 architecture…Read more

The Iranian state-sponsored threat actor MuddyWater is conducting “false flag” operations by masquerading as the Chaos ransomware-as-a-service group. These attacks utilize high-touch social engineering via Microsoft Teams, where attackers engage employees in interactive screen-sharing sessions to harvest credentials and manipulate multi-factor authentication. While the group adopts the branding and negotiation tactics of cybercriminals, they often forgo file encryption in favor of data exfiltration and…Read more

OpenAI has collaborated with major hardware vendors including AMD, Broadcom, Intel, and Nvidia to introduce the Multipath Reliable Connection (MRC) protocol. This new open-standard networking technology is designed to prevent congestion and hardware failures from stalling massive GPU clusters during AI training. By contributing the protocol to the Open Compute Project (OCP), the partners aim to establish a specialized Ethernet-based architecture that rivals the performance of proprietary…Read more

A sophisticated Linux malware platform named Quasar Linux, or QLNX, is currently targeting DevOps environments and software developers. This modular threat combines remote access trojan capabilities with rootkit functionality to maintain a persistent and stealthy presence on infected systems. Security researchers have observed the malware focusing on workstations and servers that interact with critical infrastructure like GitHub, AWS, Docker, and Kubernetes.

To evade detection, QLNX operates primarily…Read more

Google is reportedly developing an autonomous AI agent codenamed Remy, designed to transform the Gemini app into a proactive personal assistant. Currently in an internal “dogfooding” phase, the tool is being tested within a staff-only version of Gemini to handle multi-step workflows across work and daily life. Unlike standard chatbots that primarily respond to prompts, Remy is built to execute actions on a user’s behalf, such as managing calendars, summarizing emails, and monitoring specific digital…Read more

Microsoft is introducing a new security feature called Brand Impersonation Protection to combat the rising threat of voice-based phishing and social engineering. This tool specifically targets inbound Voice over IP (VoIP) calls from external contacts who have not previously interacted with the organization. By leveraging Microsoft’s threat intelligence, the system identifies callers attempting to pose as trusted brands or government agencies to steal sensitive corporate data.

The feature functions as an…Read more

AMD is implementing a new hardware feature called Highest Frequency to improve how operating systems manage processor performance and workload distribution. This capability is part of the Collaborative Processor Performance Control (CPPC) framework, which allows the hardware to communicate its performance capabilities directly to the system scheduler. By introducing a dedicated register, AMD aims to replace the current method where Windows and Linux estimate maximum boost clocks through mathematical…Read more

Microsoft is undergoing a significant leadership reshuffle following the retirement of Rajesh Jha, a 35-year veteran who led the Experiences and Devices group. LinkedIn CEO Ryan Roslansky is expanding his responsibilities to lead a newly formed Work Experiences Group, which now includes Microsoft Teams and the Office suite. This move aims to unify professional networking with productivity tools, further integrating the “economic graph” of LinkedIn with the Microsoft 365 ecosystem.

Charles Lamanna has been promoted to…Read more

The United States Department of Defense has expanded its list of approved artificial intelligence suppliers to include Microsoft, Amazon, Nvidia, and the startup Reflection AI. These vendors join existing partners like OpenAI and Google to provide generative AI capabilities for highly sensitive operations involving secret and top-secret data. The agreements aim to prevent vendor lock-in and ensure the military maintains long-term flexibility by diversifying its technological base across multiple frontier AI…Read more

AMD has submitted a new Linux kernel patch series to support a “HighestFreq” register within the Collaborative Processor Performance Control (CPPC) framework. This hardware-level addition addresses a long-standing limitation where operating systems had to estimate peak CPU boost clocks using linear interpolation, which often resulted in inaccuracies across different cores. By allowing firmware to directly expose the actual maximum frequency to the OS, the update ensures more precise CPU capacity…Read more

Palo Alto Networks has issued an urgent advisory regarding a critical zero-day vulnerability, tracked as CVE-2026-0300, which is currently under active exploitation. The flaw is a buffer overflow within the User-ID Authentication Portal, a feature used to identify users on the network when automatic mapping is unavailable. This security weakness allows unauthenticated remote attackers to execute arbitrary code with full root privileges by sending specially crafted network packets to the device.

The…Read more

Anthropic has introduced ten ready-to-run agent templates designed to automate high-volume tasks within the financial services and insurance sectors. These templates function as reference architectures that combine specific task instructions, governed data connectors, and specialized sub-agents to handle complex workflows like KYC screening and month-end closing. Administrators can deploy these tools as plugins within Claude Cowork and Claude Code or as “cookbooks” for autonomous execution via the Claude…Read more

Amazon Web Services has launched AI Traffic Analysis dashboards for AWS WAF to help administrators identify and manage the increasing volume of automated agents. This feature integrates with the AWS WAF Bot Control managed rule group, which has expanded its detection catalog to track over 650 unique AI bots and research scrapers. The tool addresses the growing challenge of AI agents accounting for up to 60% of web traffic, often incurring high infrastructure costs without providing direct business…Read more

The vendor cited…Read more

Researchers at EPFL have introduced Synthegy, a framework that leverages large language models to assist in the complex process of molecular design. Retrosynthesis is a chemical strategy where scientists work backward from a desired final molecule to identify the necessary starting materials and reaction steps. While traditional computational tools can map out vast numbers of potential chemical pathways, they often lack the strategic intuition required to select the most efficient or realistic…Read more

Mistral AI has introduced Mistral Medium 3.5, a unified dense model designed to handle complex reasoning, coding, and agentic tasks within a single architecture. This release replaces several specialized predecessors and powers the updated Vibe platform, which now supports remote, asynchronous coding sessions. The model is available via API or can be self-hosted on hardware with at least four GPUs, utilizing a modified MIT license for commercial use.

The new cloud-based agents allow developers to offload…Read more

GitHub has announced the general availability of secret scanning within its Model Context Protocol (MCP) server. The MCP is an open standard that allows AI coding agents and IDEs, such as Visual Studio Code and GitHub Copilot, to securely access external tools and data. This integration enables developers to scan code for exposed credentials using natural language prompts before committing changes or opening pull requests.

The production release now fully honors existing push protection settings configured at the…Read more

GitHub has launched a public preview of dependency scanning for its Model Context Protocol (MCP) server, enabling AI agents to identify vulnerable packages before code is committed. The Model Context Protocol is an open standard that allows AI assistants to securely interact with external data sources and tools like the GitHub Advisory Database. This new capability integrates directly into MCP-compatible environments, including Visual Studio Code and the GitHub Copilot CLI, to provide real-time security…Read more

A sophisticated Linux implant named Quasar Linux (QLNX) is targeting development and DevOps environments to facilitate supply chain attacks. The malware specifically compromises systems running npm, PyPI, GitHub, and AWS to harvest sensitive cloud credentials and SSH keys. By gaining a foothold on developer workstations, threat actors can bypass traditional enterprise security perimeters and inject malicious code directly into software delivery pipelines.

The toolkit employs a dual-layer stealth…Read more

OpenAI has launched official add-ons for Microsoft Excel and Google Sheets, allowing users to perform complex data tasks using natural language. This integration enables the creation of entire workbooks from scratch, the generation of complex formulas, and the automated cleaning of messy datasets. The tool operates within a sidebar, providing a conversational interface that can explain its reasoning and highlight specific cell references for transparency.

The service is powered by the new GPT-5.5 Instant model,…Read more

Microsoft has officially acknowledged the performance issues plaguing the Windows 11 File Explorer, which stem from layering modern WinUI 3 and XAML frameworks over legacy Win32 code. To address sluggish launch times, the company introduced a background preloading feature that caches essential application components in memory before the user initiates a launch. While this “startup boost” technique reduces initial wait times, it increases idle memory usage by approximately 35MB and does not resolve…Read more