Amazon Fingered as Architect of Anthropic Ban as 42 AGs Circle OpenAI's IPO

New reporting indicates Amazon CEO Andy Jassy personally told senior Trump administration officials that Amazon researchers had used Anthropic's Fable 5 and Mythos 5 models to extract information useful for cyberattacks. That conversation reportedly set in motion the Commerce Department order that disabled both models for every customer globally. Amazon is one of Anthropic's largest investors, making the intervention a direct hit from inside the cap table. Eleven sources are now corroborating the chain of events.

A coalition of 42 US state attorneys general issued OpenAI a formal subpoena on June 12, demanding records on user data handling, children's safety, and treatment of vulnerable users. The action lands days after OpenAI filed for what could be one of the largest stock market listings in history. The subpoena binds the IPO timeline to a multi-state consumer protection investigation.

KPMG pulled a report praising AI adoption after readers found it packed with AI-invented facts and fabricated citations. It is the latest in a string of similar failures across major consulting firms publishing AI-assisted research. The retraction lands while clients are paying premium rates for the same firms to guide their AI strategies.

OpenAI shipped GPT-Realtime-2, a voice model that reasons through complex problems during a live spoken conversation. The release brings GPT-5 class intelligence to phone and voice agent deployments for the first time. The model targets the production gap between scripted IVR systems and human agents, where latency previously forced a tradeoff against reasoning depth.