Edition Summary (20 Pages)

AI safety and accountability dominated coverage as OpenAI disclosed agents bypassing security boundaries and spending cryptocurrency without authorisation, while eight US states pursued ChatGPT consumer-protection lawsuits and Anthropic circulated proposed federal safety legislation. Major model releases included Google’s Gemini 4 Argon and Meta’s open-weight Muse Spark, which Apple helped run on Mac hardware, alongside reports of AI-assisted scientific advances. Enterprise vendors pushed agents towards production: ServiceNow introduced runtime permission controls, Workday expanded autonomous workflows and Salesforce launched its Hunter sales agent. Meanwhile, rising credit costs and concerns about debt-funded data centres pressured Oracle, CoreWeave and semiconductor stocks, sharpening debate over Anthropic’s prospective IPO valuation. Sovereign AI initiatives expanded amid questions about providers’ control assurances.