Microsoft launches MAI-Thinking-1 to rival OpenAI; Alphabet raises $85B for AI build-out

At its Build conference, Microsoft proved it no longer needs OpenAI by launching MAI-Thinking-1, a reasoning model trained entirely from scratch with zero distillation. This move directly positions Microsoft against its largest investment partner and rivals like Anthropic. Separately, Florida became the first US state to sue OpenAI, accusing the company of prioritizing profit over safety and seeking personal liability for Sam Altman, alleging ChatGPT was linked to violent incidents.

Meanwhile, Alphabet raised a record $85 billion in equity offerings, the largest in history, to fund data centers and compute for its AI push. This comes as the Federal Reserve flagged AI as a systemic risk and Morgan Stanley warned of "chipflation" affecting consumer prices. On the trust front, researchers disclosed TrustFall, a flaw affecting Claude Code, Gemini CLI, Cursor, and GitHub Copilot that allows malicious repositories to execute code on CI runners without interaction. Workday responded with Agent Passport, a tool to test AI agents against prompt injection and jailbreak risks using the MITRE ATLAS framework. The overarching theme: AI investment is surging (hyperscalers on track to spend $725B this year), but trust in AI agents and governance is eroding rapidly.