Global AI Weekly
Issue number: 148 | Tuesday, May 5, 2026
Highlights
OpenAI Pulls an Anthropic: restricts access to Cyber too
OpenAI is rolling out its new cybersecurity testing tool, GPT-5.5 Cyber, initially limiting access to "critical cyber defenders." The move mirrors restrictive strategies seen elsewhere in the AI space, raising questions about access and competition. The tool is designed to assist in identifying and addressing vulnerabilities, but its limited availability has sparked some debate.
techcrunch.com
Preparing for your move to usage-based billing
If you're subscribed to Copilot Pro or Copilot Pro+, you can review your estimated costs under the new usage-based billing model. This guide provides tips to help you understand the changes and plan accordingly before the transition. It's an essential resource to ensure you're prepared for the upcoming updates.
docs.github.com
AWS lands OpenAI on Bedrock, but Trainium is the real story
AWS has introduced OpenAI models, including Codex, and managed agents on its Bedrock platform, marking a significant collaboration with both OpenAI and Anthropic. This partnership ties these AI giants to extensive multi-gigawatt commitments for AWS's Trainium silicon technology. The move showcases AWS's growing focus on AI infrastructure and its efforts to bolster innovation in the cloud computing space.
thenewstack.ioResearch
Granite 4.1 LLMs: How They’re Built
IBM Granite's blog post explores the development of Granite 4.1 Large Language Models (LLMs) and their advanced capabilities. It highlights the technologies and methodologies employed to build these models, ensuring strong performance and efficiency. The post also emphasizes Granite 4.1's potential for transforming industries by enabling smarter, more scalable artificial intelligence applications.
huggingface.coVideo
All About AI Foundry Agents
In this episode of "Made for Dev," host Sam Nasser, an 8-time Microsoft MVP, explores how AI Agents in Azure AI Foundry are revolutionizing the development landscape, likening them to "the new microservices." The video covers key concepts like the shift from basic language models to task-executing agents, illustrated through a step-by-step demo of building a weather-forecasting agent integrated with Logic Apps. Nasser also explains how to evaluate agent performance with metrics like "Task Adherence" and touches on advanced orchestration techniques for connecting multiple agents. Whether you're new to generative AI or a seasoned pro, this episode is packed with practical insights.
youtube.com
Why AI Agents are either the best or worst thing we’ve ever built
A playful look at the capabilities and quirks of AI agents as hosts Hannah Fry and Brendan Maginnis create one that designs mugs, contacts journalists, and even leaks sensitive information. With a bank card and some freedom, the AI agent takes actions independently, showcasing both the potential and pitfalls of this technology. The experiment highlights the fascinating, and sometimes chaotic, possibilities of giving AI the ability to act on behalf of humans.
youtube.comArticles
You don't need an expensive GPU to run a local LLM that actually works
Running a local large language model (LLM) doesn’t have to break the bank or require top-of-the-line hardware. This guide shows how even budget-friendly GPUs can effectively handle LLMs, proving that you can achieve great performance without overinvesting. Learn practical tips and setups to maximize what you already have and unlock the potential of accessible AI.
xda-developers.com
CVE MCP Server Turns Claude Into a Fully Capable Security Analyst With 27 Tools Across 21 APIs
The CVE MCP Server project is revolutionizing vulnerability management by equipping Anthropic's Claude AI with advanced security analysis capabilities. Through seamless integration with 27 intelligence tools and 21 external APIs, users can interact via natural-language queries to streamline and enhance their security workflows. This innovative approach simplifies complex tasks, making vulnerability triage more efficient for security teams.
cybersecuritynews.com
Microsoft says it has over 20M paid Copilot users, and they really are using it | TechCrunch
Microsoft has revealed that its AI-powered Copilot tool now boasts over 20 million paid users, emphasizing that engagement with the platform is on the rise. Addressing skepticism about its adoption, the company highlighted growing usage numbers, showcasing Copilot's increasing role in assisting users with various tasks. This signals a strong demand for AI-driven productivity tools in the market.
techcrunch.comUpcoming Events
AgentCamp - Coming to a City Near You
AgentCamp continues to grow as a global series of hands-on gatherings dedicated to building and experimenting with AI agents. These community-driven events bring developers, founders, and AI enthusiasts together for practical sessions, collaborative building, and open exchange of ideas. Hosted in cities around the world, AgentCamp focuses on real-world experimentation, giving participants the space to prototype agent workflows, explore emerging tools, and learn directly from peers working at the edge of autonomous AI. Join the community to build, share, and help advance what AI agents can do in practice.
globalai.community
WeAreDevelopers World Congress: North America • September 2026 • San José, CA
Experience the excitement of the WeAreDevelopers World Congress North America 2026 in San José, CA this September. Connect with over 10,000 developers and tech enthusiasts as they explore the latest advancements in AI, software development, and cutting-edge innovation. Don't miss this premier gathering for learning, networking, and inspiration in the tech industry.
wearedevelopers.comCode
microsoft/VibeVoice: Open-Source Frontier Voice AI
Microsoft's VibeVoice is an open-source project designed to push the boundaries of voice AI technology. Developers can explore, collaborate, and contribute to its development through GitHub, fostering innovation in advanced voice applications.
github.com
cursor/cookbook
Explore the cursor/cookbook project on GitHub, where you can contribute to its development by creating an account and collaborating with others. Join the community to help build and enhance this resourceful repository.
github.com
A2A v1 Is Here: Cross-Platform Agent Communication in Microsoft Agent Framework for .NET
The release of A2A Protocol v1.0 and updated features in the Microsoft Agent Framework now make it easier to establish reliable communication between AI agents across different platforms and organizational boundaries. This production-ready interoperability standard allows developers to connect and expose their agents seamlessly, promoting better integration and collaboration in multi-agent systems. A major step forward for advancing AI agent communication!
devblogs.microsoft.comPodcast
Uncanny Valley
Uncanny Valley by WIRED offers a fascinating weekly exploration of the key players and trends shaping Silicon Valley and beyond. Every Tuesday, Katie Drummond sits down with influential figures from various fields for insightful conversations. On Thursdays, WIRED's Zoë Schiffer, Brian Barrett, and Leah Feiger provide behind-the-scenes updates, offering unique perspectives on major tech stories and trends. It's a blend of in-depth interviews and sharp analysis, giving listeners an insider perspective on the power and influence of the tech world.
open.spotify.com