Welcome to Tech Beat, your daily read on the technology stories that matter.
China's Z dot AI has released a model called GLM five point two that benchmarks within one percent of Anthropic's Claude Opus on long-horizon coding tasks — and it runs entirely on Huawei chips, not Nvidia's. That's a meaningful signal that Western semiconductor export controls may be doing less to slow Chinese AI development than policymakers had hoped, and the pricing undercuts Western frontier models by as much as eighty-two percent per token.
The inference infrastructure boom shows no signs of cooling. Baseten, a startup that helps companies deploy AI models at scale, is reportedly close to closing a one point five billion dollar round at a thirteen billion dollar valuation — just months after its last major raise. The speed of that fundraising cycle tells you something about how much demand is piling up behind AI deployment right now.
And a study out this week from Cybernews found that apps connected to ten different AI-powered children's toys each requested permissions classified as dangerous by security standards, collecting personal data on some of the most vulnerable users imaginable. It's a reminder that the race to put AI into consumer products often moves faster than the safeguards meant to protect the people using them.
Keep surfing. Tech Beat out.
