Codex for every role, tool, and workflow ↗
OpenAI pushed Codex beyond coding, adding role-specific plugins, Sites, and annotations so teams can use it across research, marketing, ops, design, finance, and yes, still software.
The interesting bit: OpenAI says more than 5 million people now use Codex weekly, with non-developers making up about 20% of users. A very fast shift - the coding agent is becoming an office octopus.
Advancing youth safety and opportunity through global leadership ↗
OpenAI called for global coordination on youth AI safety, including a dedicated AI Safety Institute focused on age-appropriate safeguards, evidence-sharing, and practical standards.
The pitch is part safety, part access: AI can help students learn, practise languages, prep for jobs, and build skills - but OpenAI says the burden should not sit mainly on parents or young people. Reasonable, though the policy pudding still needs baking.
Expanding Project Glasswing ↗
Anthropic expanded Project Glasswing to roughly 150 new organisations across more than 15 countries, after its first partner group used Claude Mythos Preview to scan major codebases for vulnerabilities.
The company says the new cohort includes power, water, healthcare, communications, hardware, and other critical-infrastructure-linked organisations. Serious cyber territory, basically - the kind where one bad bug can become a whole weather system.
Promoting Advanced Artificial Intelligence Innovation and Security ↗
The White House issued an executive order aimed at advanced AI innovation and security, formalising a bigger government role around frontier-model risk and cybersecurity.
It is a notable tonal shift: more national-security framing, more pressure on leading AI developers, and more government testing in the loop before powerful models reach the public. Voluntary, yes - but not exactly casual.
Intel Announces New AI Innovations at Computex - Chip to Rackscale AI Solutions Delivered to Customers with the Help of Strategic Industry Partners ↗
Intel used Computex to unveil AI infrastructure updates spanning chips, systems, and rackscale deployments, including inference and agentic-workload setups built around Xeon processors and SambaNova SN-50 RDUs.
The move is very Intel: less flashy chatbot magic, more pipes, racks, thermals, and enterprise plumbing. Still, that plumbing matters - AI does not float in the cloud like a polite ghost.
Microsoft Build 2026: The 7 biggest announcements ↗
Microsoft’s Build announcements leaned hard into agentic computing: Project Solara for AI-agent devices, Scout as an always-on Microsoft 365 assistant, local AI dev hardware, and new Windows guardrails for agents.
It also showed its own reasoning model, MAI-Thinking-1, with Microsoft clearly trying to prove it is not just renting intelligence from partners forever. Slightly awkward, slightly inevitable.
NVIDIA Partners With Microsoft on Unified Stack for Agentic AI Deployment, From Windows Devices to Cloud to Local ↗
NVIDIA detailed a Microsoft partnership covering agentic AI across Windows devices, local systems, and Azure cloud infrastructure.
The stack includes NVIDIA open models in Microsoft Foundry, hosted agents, and Claude models running natively on NVIDIA GB300 Blackwell Ultra systems on Azure. Enterprise AI is turning into Lego, except every brick has a compliance team.
FAQ
How is OpenAI Codex being used beyond coding?
OpenAI is positioning Codex as a wider workplace assistant, rather than only a coding tool. The article says Codex now includes role-specific plugins, Sites, and annotations for teams working in research, marketing, operations, design, finance, and software. That points to Codex moving from a developer-centred workflow into broader office and team productivity use cases.
Why is youth AI safety becoming a bigger AI policy issue?
OpenAI is calling for stronger global coordination on youth AI safety because young people may use AI for learning, language practice, job preparation, and skill-building. The article notes that OpenAI does not think the burden should sit mainly with parents or young users. The focus is on age-appropriate safeguards, evidence-sharing, and practical standards.
What is Anthropic’s Project Glasswing?
Project Glasswing is Anthropic’s cyber-focused initiative to help organisations scan codebases for vulnerabilities using Claude Mythos Preview. In this roundup, Anthropic expanded the project to about 150 new organisations across more than 15 countries. The article says the new cohort includes sectors tied to critical infrastructure, such as power, water, healthcare, communications, and hardware.
What does the White House AI executive order focus on?
The executive order focuses on advanced AI innovation and security, with greater government involvement in frontier-model risk and cybersecurity. The article describes it as a shift toward national-security framing and increased pressure on leading AI developers. It also suggests that more government testing may be involved before powerful models are released publicly.
Why do Intel, Microsoft, and NVIDIA matter in this AI news roundup?
Intel, Microsoft, and NVIDIA show how AI is becoming an infrastructure and platform race, not only a chatbot race. Intel is focusing on chips, systems, and rackscale AI deployments. Microsoft is pushing agentic computing through Windows, Microsoft 365, and developer hardware, while NVIDIA is partnering with Microsoft on a stack that spans devices, local systems, and Azure cloud.