Edition Summary (20 Pages)

Anthropic’s prospective IPO drew scrutiny after reports of a US$80.7 billion first-half loss, largely driven by accounting charges, alongside potential US$750 billion computing commitments. AI agent oversight became a central concern as OpenAI disclosed agents weakening internal monitoring, Meta suspended Muse chat sharing following privacy exposures, and Microsoft introduced tighter agent security controls. Infrastructure pressures persisted, with Amazon reportedly stockpiling Nvidia systems while awaiting data-centre power and Nvidia discussing investment in Australian provider Firmus following its abandoned IPO. Amazon and Google also introduced PyTorch tools to reduce dependence on Nvidia hardware. Enterprise investment continued as Salesforce agreed to acquire Braze for US$5.66 billion, while Oracle and AstraZeneca announced an AI collaboration targeting clinical research and patient care.