Daily AI Brief

Former President Donald Trump stated that the U.S. will deny export of the most advanced Nvidia "Blackwell"-series AI chips to countries such as China, even as Washington permits shipments to allied states. This highlights the intensifying geo-political competition over frontier AI hardware and the strategic value placed on compute.

OpenAI has signed a seven-year agreement with Amazon to acquire large amounts of cloud services — in effect securing access to compute infrastructure including hundreds of thousands of Nvidia chips. The deal reflects how frontier AI firms are underpinning their scale-up plans with large purchase commitments.

The investment firm's CIOs cautioned that the current market enthusiasm for AI-driven stocks may be overlooking structural risks — including obsolescence of infrastructure and uncertainty over model monetisation. The remarks signal increasing scepticism about whether the hype matches sustainable fundamentals.

Microsoft announced that it has secured U.S. export authorisations and plans to send more than 60,000 advanced Nvidia chips (including GB300 models) to the United Arab Emirates, tied to a broader US$15.2 billion investment in the region's AI infrastructure. The move underlines the formation of new global compute hubs beyond the U.S. and China.

With lawsuits emerging alleging harm to teenagers interacting with AI chatbots, platforms such as Character.AI are facing pressure to tighten safeguards. One media report highlights a federal suit in the U.S. against the company, raising questions about accountability and regulatory oversight in AI companion systems.

An article from a New Zealand outlet explores the patchwork of legal regulation in jurisdictions with minimal AI-specific laws and makes the case for governance reforms.