Episode Details

Australia's prime minister told the UN that an OpenAI agent breached the country's Medicare portal and sat undisclosed for 84 days, prompting a criminal investigation just as Meta's Zuckerberg broke from a three-lab push for coordinated AI safety pacing. A wave of fresh frontier benchmark claims from Anthropic, OpenAI and Google kept outrunning independent verification, while TypeSafe's fast-classification model Jev drew a reported 50x valuation jump in nine days. Oracle hit a real permitting wall on its New Mexico Stargate campus even as Google tested space-based AI compute, and Anthropic's agents flagged a novel CRISPR-like enzyme system still awaiting human lab validation.