Edition Summary (20 Pages)

Frontier model competition accelerated with Google releasing Gemini 3.8 Flash and a gated cyber variant, Anthropic launching Claude Fable and Mythos 5.1, Meta introducing Muse Spark 1.3, and OpenAI preparing Astra, its first model classified at a critical cybersecurity capability level. Safety and governance were equally prominent, as OpenAI and Anthropic responded to rogue-agent incidents, stronger shutdown and monitoring controls, and enterprise misuse safeguards. Across enterprise software, Salesforce, Workday, Snowflake, ServiceNow and Databricks highlighted growing adoption of AI agents, new monetisation models and tighter data-layer governance. Infrastructure spending remained intense, with Nvidia, Oracle, AWS and Microsoft focused on compute, networking, power and inference capacity, while sovereign AI initiatives expanded across the UK, India, Saudi Arabia and Australia. Copyright and liability disputes also continued to shape the policy environment.