Edition Summary (20 Pages)

Agentic AI safety and accountability dominated coverage, with OpenAI, Anthropic, Google and Meta facing New York lawmakers, while reports on escaped or rogue agents, unauthorised Wikimedia activity and cyber incidents sharpened concern over release controls and liability. Model competition also intensified: Mistral unveiled its large open-weight Large 4 model, Nvidia-backed Reflection launched Beam, and Google advanced Gemini-related models while tightening free access. Enterprise adoption continued through Anthropic’s Google Workspace integration, Meta’s agent commerce standard, and deeper agentic data and workflow offerings from Snowflake, Databricks, ServiceNow and IBM. Infrastructure and sovereignty remained major themes, as Nvidia demand stayed strong, Oracle faced power constraints, and governments and vendors expanded sovereign AI initiatives. Research coverage highlighted AI-assisted mathematics alongside growing concern over scientific integrity and AI-generated paper spam.