The morning briefing.

AI Policy & Geopolitics. The US government's export control order forced Anthropic to suspend its Fable 5 and Mythos 5 models globally, reportedly after security research from Amazon raised concerns with the White House. This move has significant implications for frontier model development and international AI collaboration. In response, China's Zhipu AI open-sourced its GLM-5.2 model with a 1 million token context, directly citing US export restrictions as a catalyst. Meanwhile, the EU continues its push for cloud autonomy and open-source adoption, adding new layers of complexity for tech buyers.

AI Adoption & Trust. Trust in AI adoption faces new challenges as KPMG pulled a report containing fabricated AI case studies involving major organizations like UBS and the NHS. This incident highlights the risk of "secondary hallucinations" from trusted consulting firms, undermining confidence in AI's business value. Despite this, AI agent usage soared 90% in the past year, yet nearly half of workers still fear AI-driven job displacement. Microsoft is proceeding with a large-scale deployment of Copilot AI tools to over half a million NHS England staff, aiming to improve service delivery and reduce costs.

Hardware & Supply Chain. The global tech supply chain is experiencing significant shifts, with surging DRAM and NAND prices forcing smartphone makers to downgrade entry-level specs from 12GB to 6GB RAM. This trend is expected to continue for up to two more years, impacting consumer electronics. Concurrently, 75 data center projects were disrupted in Q1 2026 due to increasing grassroots opposition, indicating growing challenges for infrastructure expansion. Glass core substrates are a hot topic in advanced semiconductor packaging, though their commercialization timeline remains debated.

AI Agent Evolution. The evolution of AI agents is accelerating, with major platforms integrating intelligent cores. Huawei unveiled HarmonyOS 7, transforming its Xiaoyi assistant into a system-level AI agent with over 2,100 capabilities, shifting interaction from app-centric to intent-driven. WeChat is also testing its own built-in AI Agent, prompting debate on whether this is innovation or platform defense. A new study reveals AI coding agents, while good at finding the right file, often miss critical lines of code, highlighting limitations.