by Agmo Studio
Today's digest covers a Digital Ministry plan for an AI-assisted national command centre, plus three global developments: independent evaluators gaining deeper access at Anthropic, a new open-weight model from Meta and a shopping tool from Google.
Malaysia
Digital Ministry plans National Command Centre for AI-assisted decisions
According to The Star, the Digital Ministry said in a written parliamentary reply that it plans a National Command Centre pooling data from government agencies for real-time monitoring, early risk detection and AI-assisted decision-making. It forms part of the Malaysia AI City Initiative. Phase One, running 2026 to 2027, will trial proof-of-concept projects in Kuala Lumpur, Putrajaya and nearby areas covering traffic, public health and urban sustainability. A National AI Marketplace is also planned as a sandbox for testing models on approved datasets. The ministry said AI would only support decisions, with agencies keeping final authority.
Why it matters: Businesses working in mobility, health or urban services may see future demand tied to shared government data and a model-testing sandbox. Details on participation are not yet public, so watch for follow-up announcements.
Source: The Star
Global
Anthropic gives METR researchers employee-level access
TechCrunch reports that Anthropic has given six researchers from the nonprofit METR access comparable to its own staff, including laptops, office space and internal tools, so they can independently assess its models and risk practices. METR will publish its own findings, and Anthropic has said it will not block publication, though it may require redactions in narrow categories such as customer data. Anthropic is also building a tiered system of outside evaluators. According to the report, OpenAI pledged a similar arrangement soon after the announcement.
Why it matters: Independent evaluation of frontier models may become a common reference point for vendor risk reviews. Malaysian firms buying AI services can ask providers what outside testing their models have undergone.
Source: TechCrunch
Meta releases open-weight model that can run on high-end Macs
TechCrunch reports that Meta has released Muse Spark Open, an open-weight model with 816 billion parameters and a one million token context window, aimed at coding and agentic tasks. Meta says it beats other popular open-weight models, based on evaluation results it published. Running it locally reportedly needs at least 1 terabyte of RAM, with Mac Studio systems using the M5 Ultra chip named as the best-performing setup. Meta is also offering the model through several API platforms.
Why it matters: Open-weight models give companies an alternative to closed services, which can matter for data control and cost. The hardware needs suggest most Malaysian firms would use hosted access rather than run it in-house.
Source: TechCrunch
Google launches Universal Cart shopping tool in the US
According to TechCrunch, Google has made its Universal Cart available to US users after previewing it at its May developer conference. The tool collects products found through Google Search and the Gemini app, keeps shopping progress across visits and can gather items from multiple retailers. It can send alerts on price drops or restocks and consider card perks at checkout. Checkout relies on Google's Universal Commerce Protocol, an open standard meant to help AI agents and retailers work together.
Why it matters: Agent-driven shopping could change how online customers discover and buy products. Malaysian e-commerce sellers may want to track whether open commerce standards like this reach their markets and how product data is presented.
Source: TechCrunch
AI Daily News Corner is compiled with AI assistance from the sources linked above and summarised in our own words. Details can change, so check the original reports. More daily AI news · AI consulting from Agmo.
