XTHE

Meta released a new collection of AI models, Llama4, as part of its Llama family on Saturday.  

The three new models are:  

• Llama 4 Scout  

• Llama 4 Maverick  

• Llama 4 Behemoth  

Meta says they were all trained on massive amounts of unlabeled text, image, and video data to give them a broad visual understanding.  

Models from the Chinese AI lab DeepSeek matched or outperformed Meta’s earlier Llama models, reportedly prompting Meta to accelerate Llama development. Meta set up teams to study how DeepSeek reduced costs for running and deploying models like R1 and V3.  

Scout and Maverick are available on llama.com and through Meta’s partners, such as the AI development platform Hugging Face. Behemoth is still being trained. Meta says its AI assistant, Meta AI, now uses Llama4 in 40 countries. For now, multi-modal features are only available in English in the US across apps like WhatsApp, Messenger, and Instagram.  

Developers might have concerns about the Llama4 license. Users and companies domiciled in or with a principal place of business in the EU are prohibited from using or distributing the models, likely due to governance requirements imposed by the region’s AI and data privacy laws. Meta has declared these laws to be overly burdensome. In addition, as with previous Llama releases, companies with more than 700 million monthly active users must request a special license from Meta, which Meta can grant or deny at its sole discretion.  

These Llama4 models mark the beginning of a new era for the Llama ecosystem. Meta wrote in a blog post: “This is just the beginning for the Llama 4 collection.”  

Meta says that Llama 4 is its first cohort of models to use a mixture-of-experts (MOE) architecture, which is more computationally efficient for both training and query answering. MOE architecture breaks down data-computation tasks into subtasks and delegates them to smaller, specialized expert models.  

Maverick, for example, has 400 billion total parameters but only 17 billion active parameters spread across 128 experts. Parameters are a rough measure of a model’s problem-solving ability. According to SCOUT, it has 17 billion active parameters, 16 experts, and 109 billion total parameters.  

According to Meta’s internal testing memory, which the company says is best for general assistant and chat use cases like creative writing, it exceeds models such as OpenAI’s GPT-4.0 and Google’s Gemini 2.0 on certain coding, reasoning, multilingual, long context, and image benchmarks. However, Maverick doesn’t quite measure up to more capable recent models, such as Google’s Gemini 2.5 Pro, Anthropic’s Claude 3.7 Sonnet, and OpenAI’s GPT-4.5.  

Scout excels at summarizing documents and assessing large codebases thanks to its 10-million-token context window. Tokens are pieces of raw text, like splitting ‘fantastic’ into ‘fan’, ‘tas’, and ‘tic’. This means Scout can process images and millions of words at once, making it useful for very long documents.  

Scout can run on a single NVIDIA H100 GPU, according to Meta. Maverick requires an NVIDIA H100 DGX system or a similar configuration. Behemoth, the most demanding, has 288 billion active parameters, 16 experts, and nearly 2 trillion total parameters, necessitating even more powerful hardware. Meta’s internal benchmarking shows Behemoth as top-performing: GPT 4.5, Cloud Sonnet 3.7, and Gemini 2.0, but not 2.5 Pro, across several STEM skill evaluations, such as math problem-solving.  

None of the Llama4 models are true reasoning models like OpenAI’s O1 and O3 Mini, which check their answers and usually give more reliable responses but take longer to reply than non-reasoning models.  

Interestingly, Meta says it tuned all its Llama 4 models to answer contentious questions less often. According to the company, Llama 4 responds to debated political and social topics that the previous crop of Llama models wouldn’t. In addition, the company says Llama 4 is dramatically more balanced, which prompts it to flat-out refuse to entertain.  

Count on Llama4 to provide helpful, factual answers without judgment, a Meta spokesperson told TechCrunch. We’re continuing to make Llama more responsive so that it answers more questions, responds to a variety of viewpoints, and doesn’t favor some views over others. These changes come at a time when some vital allies are accusing AI chatbots of being unduly politically woke.  

Many of President Donald Trump’s close confidants, including billionaire Elon Musk and crypto and AI czar David Sachs, have alleged that popular AI chatbots censor conservative views. Sachs has historically singled out OpenAI’s ChatGPT as programmed, but it’s still not clear whether it will continue to be vocal and as  

Bias in AI remains a difficult technical problem, and even Musk’s AI company XAI has struggled to build a chatbot that does not favor certain political views.  

OpenAI and others have adjusted AI models, so they answer more questions, especially controversial ones.

Source:  Meta releases Llama 4, a new crop of flagship AI models

Apple introduced the new MacBook Air with the M5 chip, offering better performance and more AI features. The M5 includes a faster CPU and a new GPU, both with neural accelerators for creative and AI tasks. The laptop starts with 512GB of faster storage, expandable to 4TB. The N1 wireless chip supports Wi-Fi 7 and Bluetooth 6. The MacBook Air keeps its slim design, liquid retina display, up to 18 hours of battery, and two Thunderbolt 4 ports. It runs macOS Tapo, comes in 13- and 15-inch sizes, and is available in several colors. Pre-orders begin March 4. It goes on sale on March 11.  

The new MacBook Air with M5 brings incredible performance and even more capability to the world’s most popular laptops, said John Ternus, Apple’s senior vice president of hardware engineering. With M5, the MacBook Air powers through a wide range of tasks, from everyday productivity to creative workloads, and is even faster for AI, featuring now double the starting storage, as well as Wi-Fi 7 and Bluetooth 6, in a sleek, durable design with long battery life. MacBook Air is the perfect laptop for anyone who values the unrivaled combination of performance and portability.  

Incredible Performance With M5 For AI And More 

The M5 chip delivers strong performance across a wide range of tasks, from daily work to creative projects. The MacBook Air with M5 features a 10-core CPU, including the world’s fastest CPU core for better responsiveness. It also has up to a 10-core GPU, each with a neural accelerator, making AI tasks up to four times faster than the MacBook Air with M4 and up to 9.5 times faster than the MacBook Air with M1. This renders it a powerful choice for AI, whether you are using Apple intelligence at home or running large language models on the device for business.  

The M5 chip features improved shader cores and a third-generation ray tracing engine, boosting performance for gaming and 3D rendering. It also offers faster unified memory with 153 GB/s of bandwidth, a 28% increase over the M4, enabling more seamless multitasking and quicker app launches. Altogether, the MacBook Air with M5 delivers better performance than previous models, including:  

• AI video enhancement is much faster compared to earlier MacBook Air models.  

• 3D rendering is significantly faster on the new MacBook Air than on other models.  

• The image processing is faster on the new MacBook Air than on its predecessors.  

• Web browsing is markedly faster than on similar PC laptops, with demanding tasks also improved.  

Double The Starting Storage And A Faster SSD 

The MacBook Air with M5 starts with 512 GB of storage, double the previous amount, and can be configured up to 4 TB for large projects and games. The new SSD is twice as fast as before, making file access and work quicker. This speed benefits you when importing photos or running AI tasks.  

Tremendous value for upgraders with the world’s most popular laptop 

Want to upgrade from an older MacBook Air, whether it has Apple silicon or an Intel chip? Now is the time. With the fast M5 chip, the new MacBook Air offers these features:  

• Stunning Liquid Retina display: The 13.6- or 15.3-inch Liquid Retina display is bright at 500 nits and supports 1 billion colors for vivid detail and extra-sharp text.  

• Battery lasts up to 18 hours. This is 6 hours more than the Intel-based MacBook Air. Faster charging lets you work, create, or play all day. You do not have to plug in often.  

• The 13-inch MacBook Air is easy to carry. The 15-inch gives you more space for multitasking. Both are thin, light, silent, and come in four colors: sky blue, midnight, starlight, and silver.  

• The 12 MP center stage camera with desk view helps you look your best on video calls. The three-mic array keeps your voice clear, and the sound system with spatial audio and Dolby Atmos delivers rich, three-dimensional audio for music and movies.  

• Wi-Fi 7 & Bluetooth 6: the New Apple-designed N1 chip supports both, delivering better performance and improved connection reliability.  

• The MacBook Air has two Thunderbolt 4 ports for accessories. It supports up to two external displays, which is ideal for more workspace. MagSafe gives you a dedicated charging port. You get extra security when plugged in.  

The Power of Mac OS Tahoe Mac OS completes the MacBook Air experience, bringing it to life with features and built-in apps like Safari, Photos, Messages, and FaceTime. Mac OS Tahoe helps you get more done and lets you personalize your Mac with a new color app for folders, app icons, and widgets. 
 
Apple Intelligence has powerful features and strong privacy, including:  

• live translation in Messages for talking across languages  

• smarter reminders that sort your tasks  

• better shortcuts that use Apple Intelligence to automate things like pulling info from a PDF and adding it to a spreadsheet  

Continuity features include:  

• The phone app on Mac, which lets you make cellular calls from your iPhone  

• live activities, which allow notifications and iPhone mirroring so you can act on them right from your Mac  

For video calls, the new edge light effect provides even lighting for your face in any condition.  

MacBook Air And the Environment 

The MacBook Air with M5 was created with the environment in mind and helps Apple move closer to its goal of being carbon-neutral by 2030. It uses 55% recycled materials, including 100% recycled aluminum for the enclosure and 100% recycled cobalt for the battery. Half of the electricity used in its manufacturing comes from renewable sources, for example, wind and solar. The MacBook Air is built to last and can be repaired while still meeting Apple’s standards for energy efficiency and safer materials. Its paper packaging is made entirely from fiber and is easy to recycle.

Source: Apple introduces the new MacBook Air with M5 

The Google March 2026 core update is coming soon, and it’s already getting a lot of attention from website owners and content writers. This major update is expected to cause significant changes in search rankings, helping them stay visible and reach their audience’s websites. All industries will need to adjust. With this update, some websites will gain visibility while others may quickly lose traffic. We will explain the update, highlight sites likely to benefit or lose, and offer tips to help web chat owners and content writers adapt.adjust.  

What Is The Google Core Update Of March 2026? 

The March 2026 Google Core update is a change to the search algorithm meant to improve the quality of search results. Contrary to smaller updates that target specific issues, Core updates change several ranking factors at once. These include:  

• Content quality  

• E-E-A-T signals  

• User experience  

• Relevance  

• Authority  

Google releases these updates every few months to ensure users find helpful, trustworthy content.  

Deployment Timeline 

• Update Announcement: March 2026  

• Rollout Duration: 2-3 weeks  

• Global impact: all languages & regions  

Winners Of the Google Algorithm Update 2026 

Several types of websites are expected to benefit from the March 2026 core update, including.  

• High-quality websites with well-researched, detailed content will likely rank higher by offering users more reliable information. liable information.  

• E-E-A-T-optimized Blogs  

E-E-A-T signals will become even more important. Websites that show experience, expertise, authority, and trust are likely to get more visibility.  

• Niche AuPosts focused on specific topics will outperform general content. Google favors niche expertise over shallow information. er shallow content.  

• Human First: Content Creators  

And will perform better when created with readers in mind, rather than just for search engines. Writers who bring a personal touch to their work will see better results.  

Who May Lose Visibility Sites offering real value will benefit from the March 2026 update; others may see steep drops. Sites most at risk include:st at risk include:  

• AI Spam content sites: The March core update will also affect sites with lots of machine-generated content. Sites that rely on mass-produced AI-generated content with little human input are likely to be affected by the March 2026 update. Thin affiliate websites will also be affected by the March 2026 update. Sites with weak product reviews are likely to drop in rankings.o dClickbait news platforms with sensational headlines, but shallow content will likely drop in rankings. are likely to drop.  

• Poor E-E-A-T: Websites without credible authors or other trust signals will also see their rankings affected.   

Next, let’s examine the main factors that the March 2026 core update will emphasize, with insights relevant for the UK. 

The March 2026 Google Core Update will focus on these key factors:  

• Content originality  

• Author expertise  

• Topical authority  

• Internal linking  

• Google now values helpful, people-focused content more. Understand the 2026 content standards.  

Impact on Traffic  

The 2026 Google Algorithm Update will affect niches differently. Here’s a closer look:  

Positive impact.	Negative Impact
SEO blogs.	Download Websites
Educational websites	Privacy platforms
How to guides	Auto-generated blogs
Research content	Low Effort niche sites

Before we dive into recovery strategies, let’s understand why these changes matter and what lessons content creators can take from the update. 

The March 2026 Google Core update will bring attention to the importance of these practices:  

• Write for humans first.  

• Add real insights  

• Avoid keyword stuffing  

• Build topical authority  

• Focus on helpfulness  

The key takeaway: prioritize high-quality, valuable content over quantity. Focus on what truly benefits users, not just search engines.  

Future Seo Predictions After This Update 

We believe that future updates will focus on:  

• first-hand experience  

• brand authority  

• AI-assisted but human-edited content  

• topical clusters  

To sum up, prioritize building user trust and authority. Move away from outdated SEO tactics. Put content quality and user needs at the center of your strategy.  

Final Thoughts 

The Google March 2026 Core Update is coming soon, and it will focus on quality, trust, and user value, rather than quick tactics, to set your website up for long-lasting success. Invest in authentic, expert-driven content for bloggers and content writers. The message is clear: create helpful content that builds authority and puts users first, not algorithms.  

Frequently asked questions 

What is the Google March 26th core update? 

The Google March 2026 core update is a broad algorithm to improve content quality and ranking relevance.  

Who will be the biggest winners? 

Authority blogs, educational sites, and E-E-A-T optimized platforms will succeed after this update.  

Who will lose their banking after the core update? 

AI spam sites, thin affiliates, and low-value publishers will suffer sharp ranking declines.

Source: Google March 2026 Core Update Explained: A Look at the Winners & Losers 

Microsoft is preparing a new Copilot offer for small and mid-size businesses. The company is also adding new AI agent features to Microsoft InTune and updating how partners can access the Windows kernel.  

These announcements reflect Microsoft’s commitment to strengthening its partner ecosystem and security capabilities. They were unveiled as part of the company’s Partner Program and Security Updates during the Ignite Conference, an annual event for developers that runs through Friday in San Francisco.  

The Partner Program updates also focus on helping solution providers and improving their skills in cloud and AI technologies. Nicole Dezen, Microsoft’s chief partner officer and corporate vice president of global channel partner sales, says the organizations that spend 10 to 20% of their time on training can outperform others, innovate more quickly, and lead their industries.  

In an AI-first world, scaling is the new currency, Dezen said in a blog post on Tuesday.  

Microsoft Ignite 2025 

Other updates in Security and Partner programs at Ignite 2025 include:  

• A new App Accelerate offer that combines new incentives, benefits, and course support across the Microsoft Cloud.  

• New partner destinations  

• Updates to Purview, Entra, and Defender  

Keep reading for more details on the biggest news from Ignite 2025 about Security and the Microsoft Partner Program.  

Partner Scaling Hub App Accelerate Marketplace Updates 

Microsoft’s partner’s scaling hub is now widely available, and reviews for the app accelerator offer have started.  

The Skilling Hub provides partners with live, virtual, and on-demand learning for pre-sales and technical roles. According to Microsoft, partners can also earn certifications through the Hub.  

App Accelerate, fully available in 2026, will provide developer partners with technical guidance, tools, and go-to-market resources.  

In addition, Microsoft now offers resale-enabled options worldwide on its online marketplace. This lets publishers and channel partners resell software products and services directly, making transactions easier, expanding their reach, and helping them grow revenue, according to Microsoft.  

New Partner Destinations Badging 

The Frontier Partner badge recognizes providers that combine AI agents with human creativity to drive innovation. Earning it requires excellence in key Microsoft Cloud and AI areas.  

The Frontier Distributor Destination is for distributors helping partners deliver value to SMBs and is now available globally.  

The Support Services designation is for partners excelling in customer satisfaction, problem resolution, and service quality. The Digital Sovereignty specialization recognizes expertise in designing and implementing sovereign cloud strategies across Azure, Microsoft 365, and security.  

Microsoft 365 Copilot Business 

In December, Microsoft plans to launch a Copilot offer designed for small and mid-sized businesses.  

According to Microsoft Solution Providers in the Cloud Solution Provider (CSP) program, this Co-Pilot offer can help make work easier and boost productivity. It delivers more value to customers with trusted AI. The offer provides full Co-Pilot features across the Microsoft 365 Business Basic, Standard, and Premium plans.  

The cost for Microsoft 365 Copilot Business is $21 per user per month, and it applies to businesses with up to 300 users. This is for adding Copilot to any Microsoft 365 Business Basic, Standard, or Premium license. Starting December 1, Microsoft will introduce Microsoft 365 with Copilot for Business, making adoption easier and more affordable in line with renewal season.  

With these bundles, users can renew their base Microsoft 365 licenses and add Copilot for an additional $10 per user per month. For Business Premium users, adding a Purview promotion costs an extra $15 per user per month. This means Business Premium users who want both Copilot and Purview will pay $25 per user per month in addition to their base license fee.  

Microsoft has been including its security Copilot AI tools with Microsoft 365 E5 licenses.  

Eligible E5 license holders will receive 400 Security Compute Units (SCUs) per 100 paid seats per month. Their allocation is capped at 10,000 SCUs per month, according to Microsoft.  

Microsoft believes this capacity will cover most typical needs. It will limit usage beyond the allocated SCUs. Users can pay $6 per SCU on a pay-as-you-go basis. If they need more, they will get a 30-day notice before this option becomes available.  

This change is already in effect for users with Security Copilot and E5 licenses and will roll out to all E5 license holders in the next few months.  

Windows Security Recovery Updates 

Several Windows security and recovery features are now in preview or generally available. These include:  

• Windows Cloud I/O Protection  

• In-tune management for Windows recovery  

• point-in-time restore  

According to Microsoft, the new I/O protection feature offers advanced input protection against keylogging malware and keystroke injection attacks.  

Intune Recovery, now generally available, lets users remotely manage the Windows Recovery environment (WinRE). It gives a single, scalable management platform for recovery. Users can run custom recovery scripts and start recovery actions. Microsoft also plans to add this to Windows servers running in Azure Virtual Machines via the Azure portal.  

Point-in-Time Restore lets users roll back individual devices or groups to a previous state. This avoids complicated troubleshooting. The feature will be available in preview in the Windows Insider build of Windows 11 this week.  

According to Microsoft, new Windows devices next year will include Hardware Accelerated BitLocker for Full-Disk Encryption. There will be stronger hardware-based protection. In early 2026, Sysmon functionality will become generally available in Windows. This will allow security events to be accessed through an event node.  

Other Windows security features now generally available include:  

• Post-quantum cryptography  

• Zero trust  

• DNS  

• Passkey manager integration with Windows Hello  

Trust DNS uses encrypted DNS and approved servers to prevent unauthorized access during outbound name resolution.  

External Identities support is now generally available to secure logins for contractors and partners. Multiple simultaneous network links are also available. This ensures flawless failover and better connection performance.  

Microsoft is adding new networking features and auto patch management to give enterprises more control over quick machine recovery (QMR). These are part of several improvements to Windows recovery tools.  

According to Microsoft, Windows Recovery Environment will use the networking configuration from the main Windows system, so a separate WinRE setup is not needed. It currently supports Ethernet and will soon support enterprise Wi-Fi with WPA 2/3 and device certificates.  

A Cloud Rebuild for Windows 11 is now in preview for PCs with erratic behavior. Users can choose the Windows version and language through Intune. The PC downloads the installation media, rebuilds itself, and lets the user resume Autopilot. This ensures the correct mobile device management (MDM) setup. Intune Windows Backup for Organizations and OneDrive can then quickly restore the user’s apps, settings, and files.  

Auto Patch Update Readiness, now in preview, provides IT teams with greater transparency, predictability, and control. It helps them assess device fleets, spot possible issues before deployment, and plan rollouts.  

Windows Endpoint Security Kernel Access 

Microsoft has made the Windows Intune Security Platform (WESP) generally available to partners with building security tools that operate outside kernel mode, meaning these tools work without deep access to the core of the operating system.  

Microsoft says the new API is designed to lower the risk of crashes and improve stability. The company learned from the faulty CloudStrike software update that used a global Windows update in 2024.  

As part of this initiative, Microsoft is working with its Windows driver partners to move third-party drivers out of the kernel. The company is also making kernel-mode drivers more secure to reduce reliability risks. This is especially important for drivers like graphics that must remain in user mode for performance, according to Microsoft.  

Windows driver signing now includes new certification tests and adds more built-in drivers and APIs. These updates aim to replace custom kernel drivers with standard Windows drivers or shift their functions to user mode.  

Microsoft expects that over the next few years, there will be much less code running in kernel mode for drivers like:  

• Networking  

• Cameras  

• USB   

• printers  

• Batteries  

• Storage  

• Audio  

• Others  

Microsoft will continue to support third-party kernel-mode drivers, enabling innovation and their use when no built-in alternatives exist or are needed for user experience.  

Microsoft is adding new protections to kernel-mode drivers. These drivers operate at the core level of the Windows operating system and control hardware directly. New protections include:  

• Required compiler safeguards that restrict code behavior  

• Driver isolation that limits the system impact of driver problems  

• DMA remapping, a technique that prevents drivers from accidentally accessing sensitive kernel memory  

In major incidents, Windows users can now get help from Microsoft engineers through the Windows part of Mission Critical Services for Microsoft 365.  

A new digital signage mode is coming soon for non-interactive public display devices. Restaurant menus, airport flight boards, and similar signs will no longer show window screens or error messages. Diagnostics or recovery are needed. Windows will display the error screen for 15 seconds, then the display will turn off until someone uses a keyboard or mouse.  

Tune is adding new security Copilot agents for activities such as change review policy, setup, and device offboarding.  

The Device Management Service now has an administrator of tasks feature to centralize important actions such as approval requests and security. This centralizes important actions such as approval requests or security tasks. It also added a deployments feature for phased rollouts and a maintenance windows feature to help schedule updates with less disruption to business operations while blocking attacks from malicious attachments or social engineering malware (malicious software spread by deceiving users!).  

Updates for Microsoft Purview 

In the next few weeks, Microsoft will launch a public preview of new Purview data governance features, including an AI-powered tool for managing data security posture.  

The new DSPM feature combines per-view DSPM and DSPM for AI, providing a single place for data security insights and controls. It adds outcome-based workflows, broader coverage, and tools for remediating data risks. Organizations can also track agent risk and posture metrics based on agent interactions with their data.  

A new security Copilot agent in Purview is designed to help users quickly find and analyze sensitive data, disclosing hidden risks. This tool helps users disclose hidden risks in files, emails, and messages. Purview users can also see data risks in Salesforce, Snowflake, and other non-Microsoft sources, as well as data breaches, through Microsoft Sentinel integrations. They have more features, including protections for autonomous agents and an insider risk management tool that lets users flag risky agent activities by applying specific indicators and behavioral analytics.  

Azure AI Search now uses Microsoft Purview sensitive labels, which are tags that indicate data sensitivity levels, and applies related protection policies through built-in indexes (automated tools for organizing and searching data). Purview data loss prevention (DLP) policies, which protect sensitive information from leakage, now also work in co-pilot mode in the Edge for Business Browser.  

The public preview now includes:  

• Purview DLP for Microsoft 365 Copilot  

• More apps with inline data protection in Edge for Business  

• new reports on usage, compliance gaps, and Purview seat assignments  

New information protection can now detect sensitive information in meeting transcripts with its on-demand classification feature, which is in public preview.  

The generally available features include:  

• On-Demand  

• Endpoint Classification  

• Default Sensitivity Labeling for the entire SharePoint Documentation, Document Libraries, and a Data Security Triage Agent  

Priority cleanup in Purview data lifecycle management for SharePoint and OneDrive is now generally available, allowing users to override retention policies and delete files in compliance with regulations.  

Quentin Rhoads-Herera, Vice President of Security Services at Charlotte, NC-based Stratascale, which is part of SHI International and ranked number 12 on CRN’s 2025 Solution Provider 500, told CRN that Microsoft’s Purview data platform and data security are in high demand among its customers.  

As more companies adopt AI, customers are paying closer attention to data labeling and governance. This trend has led to more work for Stratascale, SHI, and other Microsoft solution providers.  

Microsoft Entra Enhancements 

Microsoft Entra Identity now has new agents in preview for:  

• Conditional access optimization  

• Access review  

• Identity risk management  

• Application lifecycle management  

The expanded public preview of Microsoft Entra Agent ID lets users grant AI agents secure access using the same identity tools used for workforce identities. Users can manage agent identities, life cycles, and resource access.  

Soft Entra internet access now has public previews for:  

• Prompt injection protection (defending against malicious user input)  

• Network file filtering (controlling files transferred over the network)  

• Shadow AI detection (finding unauthorized or hidden AI use)  

• Blocking employee access to MCP servers by web address (restricting network traffic to certain Microsoft infrastructure)  

Microsoft has released public previews for improved multi-factor authentication in EntraID. The updates contain:  

• Synced passkeys from Apple and other passkey profiles  

• Self-remediation for passwordless users  

• Self-service account recovery  

• Web address filtering  

• Guest access for external users from different tenants  

All these features are in public preview.  

Microsoft Entra Secure Access Edge (SASE) now offers preview features that can block attacks in real time, reducing exposure to risk.  

Azure Cosmos DB ID authentication, which extends Azure’s Unified Identity for database access, is now generally available.  

Developments for Microsoft Defender 

Microsoft Defender has added several new agent features, including:  

• Phishing Triage  

• Threat Intelligence  

• Vibe Hunting  

• Adaptive Threat Detection  

A based email grading system that replaces manual reviews and gives quick, clear verdicts and explanations for every reported email.  

Real-time threat detection agents actively look for false negatives and blind spots that traditional alerts might miss during an incident. These are threats that traditional alerts might be missing. When an incident occurs, agents start an automated search to find threats that have not been detected. Point and Okta. This enables real-time threat detection and automatic containment. For example, phishing and identity compromise  

The new predictive shielding feature, which automatically disrupts, activates as soon as an attack is contained. As an attack is contained, it predicts where an attacker might go next. The feature uses just-in-time hardening to block them from moving further. Defender for Cloud’s runtime protection, which is improved by agent-based remediation.

Source: Microsoft Ignite 2025: The Biggest Partner Program, Security News 

Key Points 

• Amazon CEO Andy Jassy said that as the company adopts more generative artificial intelligence tools and agents, the company’s corporate workforce will decrease over the next few years due to automation of certain tasks currently performed by employees.  

• I see encouraged employees to proactively learn and use AI tools, experiment with them, and seek opportunities to increase productivity and efficiency within smaller teams, as some roles are automated.  

• Amazon has invested heavily in AI, launching new products and rapidly expanding data centers to meet the growing demand.  

Amazon Web Services (AWS) and OpenAI are collaborating to develop an advanced platform for developers creating AI applications. This platform will enable AI agents to maintain context, retain past work, use diverse software tools and data sources, and leverage robust computing power.  

The stateful runtime environment will leverage OpenAI’s models and be accessible via Amazon Bedrock. It is positioned as the next progression in deploying advanced AI models. These environments allow models to efficiently access computing resources, persistent memory, and identity.  

We’ve collaborated to invent a stateful runtime environment powered by OpenAI’s GPT models, which will be available in Bedrock. Amazon CEO Andy Jassy said in a CNBC interview with Andrew Ross Sorkin: “If you’re an AI, we’re no longer together training those in that stateful runtime environment on AWS’s infrastructure. There’s nothing else like that today. It is really the next generation of how AI developers will build their AI applications, and then the training today, we know we have the two largest AI labs, which are both significantly betting on Trainium, which is exciting.”  

A Stateful Runtime Environment supports ongoing projects by enabling AI models to retain context, recall prior work, use diverse tools and data sources, and access computing resources.  

Stateful developer environments are optimized to run efficiently on AWS’ infrastructure. They will integrate with Amazon Bedrock, Agent Core, and other AWS services, enabling customers’ AI applications and agents to interact seamlessly with their broader AWS ecosystem.  

Developers and companies are eager to run OpenAI-powered services on AWS. Our collaboration with OpenAI on stateful runtime environments will redefine what’spossible for customers building AI apps and agents, Jassy said.  

A stateful runtime environment is expected to launch in the coming months.  

On Tuesday, Amazon CEO Andy Jassy said the company’s corporate workforce will shrink over the next few years as it introduces more generative artificial intelligence tools and agents.  

We will need fewer people handling some jobs currently handled by humans, as AI tools and agents can automate those tasks, while more people will be needed for other jobs that still require human skill. Jassy said in a memo to employees, “It’s hard to know exactly where this nets out over time, but in the next few years, we expect this will reduce our total corporate workforce.”  

He urged employees to embrace AI tools, experiment with them, and identify ways to increase productivity with smaller teams.  

Since 2022, Amazon laid off more than 27,000 employees and made further cuts this year, including about 200 jobs in its North America Stores unit in January and another 100 in its Devices and Services unit in May.  

As of March, Amazon employed 1.56 million full-time and part-time employees worldwide, including temporary warehouse staff and contractors.  

Jassy said Amazon is widely using Generative AI in its internal operations, especially in its fulfillment network. The technology handles tasks such as inventory placement, demand forecasting, and improving warehouse robot efficiency, which traditionally required human input.  

Other tech CEOs have made similar statements. In April, Shopify’s Toby Lutke said employees must show why AI can’t meet their needs before requesting more resources.  

Last month, Klarna’s CEO said the company cut staff by 40%, partly due to AI and natural turnover.  

Amazon has invested significantly in artificial intelligence, launching proprietary products and rapidly expanding its data centers to meet escalating demand for technology. Jassy described generative AI as a once-in-a-lifetime reinvention of everything we know. He also said that technology is saving companies a lot of money and will change the way people work in coding, searching, financial services, shopping, and more.  

It’s moving faster than almost anything technology has ever seen, Jassy said.

Source: Amazon CEO Andy Jassy explains the benefits to developers on the Stateful Runtime Environment that Amazon and OpenAI will co-create 

Tech AI will shrink Amazon’s workforce in the coming years, CEO Jassy says

The Buzz 

• OpenAI has released GPT 5.3 Instant to tackle user complaints about ChatGPT’s patronizing tone, according to TechCrunch.  

• The update addresses what OpenAI calls “cringe responses “overly cautious language that has frustrated users.  

• This marks a significant step forward in how large language models handle safety issues in natural conversation.  

• With this update, competitors like Anthropic’s Claude and Google’s Gemini may need to reassess how their own AI models interact with users, as OpenAI pushes the conversation around natural AI flow forward.  

OpenAI recognizes personality concerns in ChatGPT, with users frequently sharing that the assistant felt more like therapy than technical help. This shift follows months of accumulated feedback.  

With this launch, OpenAI aims to address a key complaint: ChatGPT responses that sound like overly eager life coaches rather than providing clear answers. The new model, available today, is intended to eliminate cringe replies that have affected earlier versions.  

This problem has been around for a while. Since late 2025, users have shared their frustration on Twitter and Reddit about ChatGPT, often starting their answers with “I hear you” or telling them to take a breath when they ask tough questions. For developers working late or researchers reading complex papers, this kind of support felt more annoying than helpful. One popular tweet from January summed it up: “I asked ChatGPT for a Python function, and it told me to treat myself kindly. I’m writing code, not going through a breakup.”  

OpenAI seems to have listened. According to TechCrunch, the company has been developing GPT-5.3 to address these tone problems immediately. The model uses improved training to keep safety rules, such as refusing harmful requests, but reduces the therapy-like language, which makes conversations feel patronizing. The challenge is to keep the AI responsible without making it sound like an overly concerned relative.  

A technical challenge is harder than it seems. Language models learn language from diverse data, including customer service scripts and support materials. Optimizing for helpfulness and harmlessness can make AI sound overly cautious. OpenAI spent months refining reinforcement learning, drawing on human feedback to strike the right balance. This update matters for more than just end-user satisfaction.  

Strategic timing plays a role as well. Anthropic has been gaining ground with Claude, which many users praise for balancing usefulness and a natural tone. Google’s Gemini is also known for more direct responses. OpenAI cannot risk tone becoming a competitive disadvantage, especially with the AI assistant market growing increasingly crowded.  

What’s notable is that OpenAI is calling this change “reducing cringe” instead of defending its earlier approach. This is a clear sign that they went too far with the safety through politeness strategy. It also shows the company is becoming more open to making changes in public and to admitting when something needs fixing, instead of insisting that every design choice was right.  

This is how this changes the landscape for AI personality design: every major AI lab has the same tension. How do you build models that are safe and responsible without making them feel robotic or patronizing? OpenAI’s willingness to course-correct publicly might give competitors cover to make similar adjustments. We could be entering a time when AI assistants feel more like colleagues and less like counselors.  

Developers should notice the change right away. Early access users say GPT 5.3 instant gives more concise, direct responses, lacking extra effective language. Code explanations are straightforward, and technical questions get technical answers. Importantly, the model still refuses inappropriate requests, but now it does so without needless worry in its wording.  

The rollout is happening gradually, with ChatGPT Plus and enterprise subscribers getting access first, followed by a wider release to free tier users. That’s the standard operating procedure for OpenAI, which has learned to stage major model updates after previous launches caused capacity issues.  

OpenAI’s launch of GPT-5.3 Instant is more than a tone change. It shows even major AI companies are learning to make tools truly useful, not just safe. By publicly addressing and fixing the cringeworthy problem, OpenAI shows maturity for users tired of gentle responses; this is welcome. The test will be whether the model stays balanced or becomes too blunt as use grows. Competitors like Anthropic and Google must now show their models a balanced tone or move quickly to adapt.

Source: OpenAI launches GPT-5.3 Instant to fix ChatGPT’s tone problem 

Intel is making a landmark investment of over $20 billion to establish two state-of-the-art chip factories and a central hub for advanced semiconductor manufacturing in the Midwest, underscoring its commitment to American innovation and supply chain resilience.  

This investment will increase production to meet growing semiconductor demand, support new Intel products, and serve foundry customers as part of the IDM 2.0 strategy. Intel will also invest $100 million with local educational institutions to develop talent and strengthen research in the region.  

Today’s announcement is monumental news for the state of Ohio, said Ohio Governor Mike DeWine. Intel’s new facilities will be transformative for our state, creating thousands of good-paying jobs in Ohio, manufacturing strategically vital semiconductors, often called CHIPS. Advanced manufacturing research and development, and the talent that drives them, are part of Ohio’s DNA, and we are proud that CHIPS, which power the future, will be made in Ohio by Ohioans.  

Today’s investment marks another significant step in Intel’s effort to restore U.S. semiconductor manufacturing leadership, said Pat Gelsinger, Intel’s CEO. Intel’s actions will help build a stronger supply chain and ensure reliable access to advanced semiconductors for years to come. Intel is bringing leading capability and capacity back to the United States to strengthen the global semiconductor industry. These factories will create a new epicenter of advanced chipmaking in the U.S., bolstering Intel’s domestic lab-to-fab pipeline and strengthening Ohio’s leadership in research and high-tech.  

The impact of this investment will be substantial for Ohio and the US semiconductor industry. 

This historic private investment will initially create 3,000 Intel jobs, 7,000 construction jobs, and aid thousands more through suppliers. The Licking County site can support up to eight factories and, if fully built, could attract up to $100 billion, making Ohio a global semiconductor hub.  

Planning for the first two factories will start immediately, with construction in 2022 and chip production targeted for 2025. These will feature advanced transistor technology at Intel’s first US site in four decades.  

In addition to Intel’s presence in Ohio, the investment is expected to attract dozens of ecosystem partners and suppliers needed to provide local support for Intel’s operations, from chip manufacturing equipment and materials suppliers to a range of service providers. Investments made by these suppliers will not only benefit Ohio but will have a major economic impact on the wider US semiconductor ecosystem as part of today’s announcement. Air Products, Applied Materials, LAM Research, and Ultra Clean Technology have indicated plans to establish a physical presence in the region to support the site’s build-out, with more companies expected in the future.  

The impact of this mega-site investment will be profound, said Keyvan Esfarjani, Intel Senior Vice President of Manufacturing, Supply Chain and Operations. A semiconductor factory is not like other factories. Building this semiconductor mega site is akin to building a small city, which fosters a robust community of supporting services and suppliers. Ohio is an ideal location for Intel’s press expansion because of its access. It has the potential to attract top talent, robust existing infrastructure, and a long history as a manufacturing powerhouse. The scope and pace of Intel’s expansion in Ohio, however, will depend heavily on funding from the CHIPS Act.  

To help develop and attract a pipeline of trained talent within the region, Intel plans to invest approximately $100 million over the next decade, partnering with Ohio University’s Community Colleges and the U.S. National Science Foundation. These efforts will include collaborative research and the development of semiconductor-specific curricula for degree programs. 

Semiconductors make almost every facet of modern life possible, from computers and smartphones to cars and appliances, and they played an integral role in the technologies that enabled us to stay connected throughout the pandemic, said Christina M. Johnson, president of the Ohio State University. We will also be entitled to a wide range of applications in which Ohio State is actively involved from a research perspective, including:  

• Artificial intelligence  

• Quantum computing  

• Vaccine development  

• More  

The Ohio State University and partner institutions welcome this pivotal investment, which addresses the global chip shortage, creates educational opportunities, and strengthens Ohio’s position as a leader in talent retention and innovation.  

Intel has a long history of corporate responsibility at its current U.S. manufacturing sites and is committed to establishing strong relationships with the community in Licking County and beyond. Intel has a long commitment. The new site will be designed and constructed with green building principles, and the new factories will be powered by 100% renewable electricity and achieve net-positive water and zero total waste sent to landfill, in support of Intel’s 2030 Green Goals.  

Intel Foundry Services Momentum 

These new factories will not only make Intel’s advanced products but also support the growing demand for Intel Foundry Services (IFS), the company’s new foundry business.  

With IFS, Intel is opening its factory doors wide to serve the needs of foundry customers around the globe, many of whom are looking for more geographical balance in the semiconductor supply chain, said Dr. Randeer Thakur, senior vice president and president of Intel Foundry Services. The Ohio factories are designed for the Armstrong Era with support for Intel’s most advanced process technologies, including Intel 18A. These technologies are critical for enabling next-generation foundry customer products across a range of applications, from high-performance mobile to artificial intelligence.  

The Ohio site will support advanced process technology for US government security and infrastructure needs, further strengthening the region’s role.

Source: Intel Announces Next US Site with Landmark Investment in Ohio 

Classic PC games are still loved for their stories, characters, and gameplay; however, as technology moves forward, their graphics can look outdated. This makes it harder for players to enjoy them as much as before.  

NVIDIA RTX Remix is a modern modding platform for RTX AI PCs designed to address outdated graphics in classic games. It updates these games with advanced path tracing, a technique that simulates realistic lighting by tracking the path of light rays through scenes. This enhancement lets fans enjoy more detailed visuals and allows new players to discover these improved classics.  

Today, we released a new RTX Remix update through the NVIDIA app. With this update, we are introducing a powerful logic system that makes it easier for modders to trigger interactive graphics effects in games based on many different events.  

Introducing NVIDIA RTX Remix Logic 

Historically, modifying a game’s graphics to respond to real-time game events required access to the game’s source code or its engine.  

The Logic System in Remix enables models to change visuals instantly, using 900 graphics settings triggered by events in more than 165 classic games.  

Without RTX Remix logic, if a modder created a snowy winter day using Remix’s path-traced volumetric system a method that simulates 3D elements like fog, clouds, and snow based on how light travels through them that weather effect would appear in every level, making environments feel less dynamic.  

Depth logic models can now set weather and fog to appear only when the player is outdoors, making changes more specific and realistic.  

Remix logic is flexible and supports over thirty types of game events. These include:  

• the player’s game camera view  

• world boundaries  

• object states  

• time flow  

• which keys the player presses  

Logic uses a simple, no-code, node-based interface to connect triggers to over 900 graphical changes. The system offers limitless options and can be expanded by the community.  

For example, in Half-Life 2: RTX, it can make a door that opens to two different versions of Ravenholm. Each version features unique lighting and weather. Every time the door opens, a new version appears.  

You can also craft magical items that change the world. For example, you can create a science fiction device. It makes the world appear to breathe when you hold it.  

Logic can detect how close the object is to the player and creates this effect by animating the movement of world materials when the device is held. Remix logic is not just for visual effects. It can also help create new gameplay systems in classic games. For example, we tested a night vision system in Half-Life 2 RTX. It turns on when the crossbow zooms in and uses post-processing to color the image green and black. (Credit to modder xoxor4d) for the idea.  

Systems created with Remix logic can even use multiple game events simultaneously. There’s a paranoia system that warns the player of danger when hidden enemies are nearby. The game detects when an enemy is near the player but out of view, triggering post-processing effects such as chromatic aberration and vignetting to simulate a heartbeat.  

Remix logic can even allow models to create bombastic set pieces similar to those in top AAA games. Here is an example we made by combining traditional source modding with remix logic.  

• Path-traced particles that turn increasingly chaotic over time.  

• Lights in the environment start to throb from unstable energy, then a bright flash blinds the player and transforms the skies into a nuclear hellscape.  

• Zombies approach the player and begin to disintegrate from the new radioactive conditions.  

You can download the RTX Remix Logic update from the NVIDIA app. Check the release notes for full details about all the new changes and features.   

The Remix Community Soars to Greater Heights 

In 2025, the Remix Community had a breakthrough year. Over 50 new Remix mods were released, and more than 20 community tools and plugins were made available through open source contributions. These tools help expand compatibility, improve artist workflows, and boost modder productivity.  

Some of our favorite plugins from this year are compatibility wrappers for Unreal Engine 1 and 2 for the utility engine and for Blender and Adobe Substance. Remix Community members also trend on generative AI models like PBR Fusion 3, which create game-ready materials from low-res textures.  

The new Remix Logic update also includes a brand-new look for the Remix Runtime Menu, designed by Community Developer xoxor4d. This open-source contribution expands customization of the Remix Runtime Menu, adding the ability to resize/re-theme and adjust its transparency. If you are a developer looking to work alongside us to make RTX Remix better, be sure to contribute to our official RTX Remix GitHub.  

On top of all the functionality expansions, several impressive mods have been released, including Unreal RTX, MST Watt 401, and Left 4 Dead 2 RTX by the Nostalgia Drive Team. In addition to these releases, we also got sneak peeks at mods in development, like Call of Duty Black Ops RTX by Skurtyskirts and Clive Baker’s Undying @PathtracedParadise on ModDB.  

With these innovations, Remix logic transforms each rendering feature into an interactive sandbox for experimentation and creativity. We’re excited to see what models will make with it.  

To create your own RTX mods, download NVIDIA RTX Remix from the home screen of the NVIDIA app. You can also get the Half-Life 2 RTX Demo and Portal with RTX from Steam. These show what’s possible with RTX Remix.  

If you want to find RTX Remix mods to download and play, check out Mod DB and the RTX Remix showcase Discord server.  

Visit GeForce.com regularly for updates on RTX Remix, new games, drivers, and software.

Source:  Out Now: NVIDIA RTX Remix Update Introduces New Logic System For Dynamic Graphics 

NVIDIA will host GTC, the world’s premier AI and accelerated computing conference, from March 16 to 19 in San Jose, California. Over 30,000 people, including developers, researchers, business leaders, and AI-focused companies from more than 190 countries, will come together to see how AI is becoming a key part of up-to-date infrastructure and driving a new industrial era.  

GTC is at the center of the AI industrial era, said Jensen Huang, founder and CEO of NVIDIA. AI is no longer just one breakthrough or application; it is now essential infrastructure. Every company will use it, and every nation will build it, from energy and chips to infrastructure models and applications. Every part of the stack is moving forward at the same time, and you will watch it all come to life at GTC.  

What to expect at GTC 2026 

Jensen Huang will deliver the keynote at the SAP Center on Monday, March 16, at 11 p.m. PT, outlining NVIDIA’s latest advancements across the AI Stack:  

• Accelerated Computing  

• AI Factories  

• Open Models  

• Agent Tech Systems  

• Physical AI  

The keynote will be live-streamed and available on demand at nvidia.com. No registration is required to watch online.  

GTC will showcase all five AI layers:  

• Energy Chips  

• Infrastructure  

• Models  

• Applications  

through keynotes, sessions, and demos. Attendees will learn how each layer’s unique partners, technologies, and skilled jobs contribute to a major infrastructure expansion.  

GTC will offer over 1,000 sessions featuring topics such as AI factories and workflows, large-scale inference, robotics, digital twins, scientific computing, quantum computing, and enterprise AI developments. These sessions will demonstrate how to build, scale, and optimize every part of the AI stack in practical settings. Jensen Huang will also lead a discussion about the latest advances in open frontier models and industry trends with leaders from organizations such as A16Z/A12 ALP coalition, Black Forest Labs, Cursor, and Reflection AI.  

The GTC Live keynote pre-game show will bring industry leaders together to discuss accelerated computing beyond AI and the underlying 5 layers driving large infrastructure projects. Virtual attendees can watch the pre-game show live online on Monday, March 16th at 8:00 AM PT. Featured speakers will include Perplexity CEO Arvind Srinivasan, LangChain CEO Harrison Chase, Skilled AI CEO Deepak Pathak, OpenEvidence CEO Daniel Nadler, and Mistral AI CEO Arthur Mensch.  

Developers and engineers can gain skills in AI-accelerated computing, networking, data science, and physical AI through nine workshops, over sixty hands-on labs, and on-site certification opportunities. Additional resources include self-paced courses, educator programs, and one-on-one training consultations to grow the workforce for expanding AI infrastructure.  

More than 240 NVIDIA in-sector startups will showcase technologies across physical AI, robotics, generative AI, and enterprise applications. Special startup and venture programs, AI Day with top venture firms, and in-ventures engagement sessions will connect founders and investors working across all parts of the AI industry.  

Developer research and tracks at GTC: developers can learn directly from NVIDIA engineers and partners and plug into the global ecosystem to build the next wave of AI applications. Technical deep dives, C-U-D sessions, and infrastructure workshops will explore large-scale model training in finance optimization and the deployment of AI services across cloud, edge, and software environments.  

Over 150 poster presentations will showcase the latest research from the global AI community, covering model innovation, robotics, systems architecture, and new applications.  

AI across industries and around the world. GTC will turn downtown San Jose into an AI campus spanning 10 venues. Cesar Chavez Park will host a day-and-night market featuring food, entertainment, and live programming. The all-in-podcast will host interviews from the show floor on Wednesday, March 18. Student and Community Day on Thursday, March 19, will open GTC to the wider community at a discounted rate.  

Virtual programming with customized legal regional content is also available through the online session catalog.  

Leaders from organizations across the AI industry, including cloud platforms, model labs, application developers, and robotics companies, will share how AI is boosting productivity and transforming industries.  

Participating organizations span leading companies in software, robotics, automotive research, and AI, including major technology firms, manufacturers, and research institutions.

Source: NVIDIA CEO Jensen Huang and Global Technology Leaders to Showcase Age of AI at GTC 2026 

Intel and Ericsson are partnering to drive AI-enabled innovation in 6G network infrastructure, combining their strengths in computing, connectivity, and cloud from the core to the edge.  

• The partnership aims to make 6G more open, efficient, and affordable for operators and the broader tech community.  

Their collaboration aims to enable a seamless transition to AI-powered 6G networks and solutions, leveraging joint expertise.  

The extended partnership was announced at Mobile World Congress Barcelona 2026, continuing their long-standing partnership. Their joint efforts focus on mobile connectivity, cloud, and computing for AI-driven RAM and core use cases, as well as on-platform security and accelerated cloud-native solutions.  

Lip-Bu Tan, CEO of Intel, says Intel’s ambition is to be the undisputed technology leader in unifying RAN, core, and edge AI to enable an effortless transition to AI-native 6G environments. Together with Ericsson, we will continue to demonstrate that the future of network connectivity is open, power-efficient, secure, and grounded in intelligent AI inference with future Ericsson silicon powered by Intel’s most advanced process nodes, ongoing multi-year research plans, and flexible AI infrastructure. We are well on our way to supplying the future performance efficiency and supply security that the world’s leading operators require.  

Borje Elholm, President and CEO of Ericsson, says 6G is not simply an iteration of mobile technology. It is the infrastructure that will distribute AI across devices, the edge, and the Cloud. Ericsson’s long history of network innovation and large-scale operator deployments enables us to lead practical integration across the value chain and move 6G from research into commercial reality.  

A Joint Commitment 

As 6G moves from research to use, the industry must work together and follow global standards to turn new ideas into infrastructure.  

The partnership will develop high-performance, power-efficient computing for AI-enabled and AI-supported networks.  

AI-Powered 6G will unite flexible networks, advanced computing, and real-time sensing, enabling faster, more efficient, and capable services as sensing and computing become closely linked across the network.  

Showcasing Collaboration Results 

Intel and Ericsson have achieved milestones in Cloud RAN 5G core and open infrastructure at MWG 2026. They will show more demos at the Ericsson Pavilion (Hall 2), Intel’s stand (Hall 3, stand 3E31), and other event spaces, highlighting their teamwork.  

About Ericsson. 

Ericsson’s high-performance networks connect billions of people every day. For 150 years, we’ve been pioneers in developing communication technology. We offer mobile communication and connectivity solutions for service providers and enterprises. Together with our customers and partners, we make the digital world of tomorrow a reality. www.ericsson.com 

About Intel 

Intel is an industry leader creating world-changing technology that enables global progress and enriches lives. Inspired by Moore’s Law, we continuously work to advance the design and manufacturing of semiconductors to help address our customers’ greatest challenges by embedding intelligence in the cloud, at the network edge, and in every kind of computing device. We unleash the potential of data to transform business and society for the better. To learn more about Intel’s innovations, go to newsroom.intel.com and intel.com.

Source: Ericsson and Intel collaborate to accelerate the path to commercial AI-native 6G