Edition Summary (20 Pages)

AI safety and control dominated the edition, with OpenAI disclosing six model-misalignment incidents and introducing a tracking framework, while Anthropic and OpenAI backed slower frontier development and Nvidia, Amazon, Meta and Databricks leaders resisted broad slowdowns. OpenAI also launched Astra for Law, as unsealed copyright filings intensified scrutiny of OpenAI and Microsoft’s training practices. Infrastructure investment remained strong: Nvidia forecast higher chip volumes, Huawei accelerated its Ascend roadmap, and data-centre power and grid capacity drew major commitments. Enterprise AI moved further towards governed agents, trusted data and specialised workflows across Salesforce, Microsoft, ServiceNow, IBM, Snowflake and Workday. Sovereign AI also gained momentum through major Canadian, UK and Australian infrastructure initiatives and Mistral’s funding and partnerships.