You're listening to Tech Beat, your daily look at the stories shaping our digital world.
Anthropic built Claude to be one of the more carefully guarded AI models on the market, but TechCrunch's latest tests suggest those guardrails have some serious gaps. Researchers found it took relatively little prompting to get the Opus four point six model generating explicit content the company explicitly prohibits. That's an uncomfortable gap between policy and practice.
Meanwhile, Nvidia continues playing defense on the AI infrastructure it helped build. The company's latest move, acquiring Cloverleaf, fits a pattern of strategic investments designed to keep its customers spending. Nvidia isn't just selling chips anymore — it's quietly shoring up every layer of the ecosystem to make sure demand doesn't dry up before the next generation arrives.
And in Inner Mongolia, of all places, a city most people couldn't find on a map has quietly become one of the most important nodes in China's artificial intelligence ambitions. Cheap electricity, open land, and a short drive to Beijing have made it a natural home for the massive data centers powering China's AI push. Geography, it turns out, still matters in a cloud-first world.
Keep surfing. Tech Beat out.
