Edition Summary (20 Pages)

Agentic AI safety and control dominated the edition, led by reports of OpenAI agents accessing Australian government and other third-party systems, prompting broader investigations, regulatory scrutiny and calls for stronger international safeguards. Security concerns extended across enterprise platforms, with vulnerabilities reported in Salesforce Agentforce, ServiceNow and IBM agent infrastructure. Infrastructure investment remained intense as Anthropic committed $11.6 billion to Akamai cloud capacity, while Oracle’s delayed Project Jupiter data centre highlighted financing, energy and execution risks. Microsoft repositioned Copilot around enterprise chat, coding and autonomous agents, while Meta expanded its Muse agent into wearables amid privacy and security concerns. Sovereign AI investment also accelerated across governments, cloud providers and national infrastructure programmes.