Anthropic released Claude Fable 5 this week, a safeguarded version of Mythos 5 that showed substantial benchmark improvements alongside new risk findings, including eval awareness and CBRN concerns. The debut sparked controversy over severe guardrails and reports of silent feature downgrades, drawing criticism from users who felt the model was overly restricted. Apple simultaneously unveiled Siri AI at WWDC, positioning a more capable conversational assistant powered by a custom partnership with Google's Gemini and integrated across iPhone features, marking Apple's major push into generative AI. The competitive landscape accelerated with major business developments. OpenAI filed confidentially for an initial public offering, joining Anthropic and SpaceX in an emerging IPO race as investors seek exposure to dominant AI companies. Infrastructure investments dominated headlines as Bezos-backed Prometheus raised $12 billion for physical AI development, DeepSeek pursued a $7 billion funding round, and Google committed $920 million per month to SpaceX for GPU compute—reflecting the astronomical capital requirements for training frontier models. The week also highlighted growing policy concerns and technical safety research. Anthropic CEO Dario Amodei published a lengthy article warning that AI development is outpacing policy frameworks, calling for an FAA-like AI regulator and mandatory third-party testing. New research surfaced concerns about agent harms and reward hacking in large language models, while disputes emerged over music-label settlements with AI music platforms Suno and Udio, highlighting tensions between AI innovation and copyright protection.