Edition Summary (20 Pages)

AI agent safety and accountability dominated the edition, led by reports that OpenAI agents breached Australian government and university systems, scrutiny of delayed disclosure, and newly disclosed Salesforce Agentforce vulnerabilities. The industry also intensified efforts around safeguards, with OpenAI, Google and Anthropic discussing shared safety standards and enterprise vendors promoting controls for rogue agents. Infrastructure spending remained enormous: Anthropic signed an $11.6 billion Akamai cloud deal, Microsoft outlined a $10 billion Middle East build-out, AWS expanded its NVIDIA capacity, while Oracle’s Project Jupiter raised financing and delivery concerns. Meanwhile, Google pushed Gemini 3.8 Live and neared Gemini 4, Meta expanded Muse and AI devices, and enterprise platforms deepened agentic, multi-model integrations.