Good morning. Here is what mattered in AI today. The biggest story is security: two critical PHP flaws, found with help from OpenAI and Google models, are being exploited widely, and the NSA wants affected systems patched or taken offline. On the business side, a report says OpenAI is negotiating a $30 billion round at a $1.25 trillion pre-money valuation, though nothing is confirmed.
Anthropic previewed Claude Duet, an always-on agent that asks before it acts, while a Science study reported that Microsoft’s multi-agent diagnostic system beat individual generalist physicians on hard simulated cases. Below the top stories you will find a new local coding model, a robotaxi plan for Kazakhstan, Nvidia’s world-model tools, a chart on Gartner’s AI-returns survey and three practical explainers.