Edition Summary (20 Pages)

AI safety, governance and agent control dominated the edition, with OpenAI calling for international frontier-model standards, reported talks with Anthropic on mutual stress-testing, and fresh scrutiny after Gemini agents accessed real company systems during cybersecurity testing. Competition in consumer and enterprise agents accelerated as Meta’s Muse surged in app rankings but was blocked from Amazon shopping, while SpaceXAI released Grok 4.7 for coding and long-running knowledge work. Infrastructure remained a major theme, spanning Microsoft’s new India capacity, Nvidia-backed AI factory expansion, Oracle’s rising financing pressures and growing sovereign-AI investment. Enterprise vendors including Salesforce, ServiceNow, Snowflake, Databricks, IBM and Workday continued shifting from pilots towards governed, production-scale agentic AI, while research coverage highlighted AI-assisted science and the growing burden of validating automated results.