Edition Summary (20 Pages)

AI safety, governance and the consequences of increasingly autonomous agents dominated the edition, with OpenAI introducing EU text watermarking, major labs testifying before New York lawmakers, and scrutiny intensifying around Meta Muse privacy, Wikimedia agent activity and AI-enabled cyber threats. Competition among model builders also shifted, as Google unveiled Gemini 4 Argon while restricting lower-tier Gemini access, Nvidia-backed Reflection introduced the open-weight Beam model, and Microsoft and Meta reduced internal use of Anthropic Claude ahead of Anthropic’s anticipated IPO. Infrastructure pressures remained prominent, spanning record-scale chip financing, data-centre power constraints and nuclear energy deals. Enterprise vendors including IBM, Salesforce, ServiceNow, Snowflake and Databricks continued pushing agentic AI into governed production workflows.