You're tuned in to Tech Beat, and here's what's moving in the world of technology today.
China's internet regulator has launched an investigation into DeepSeek and Moonshot AI following accusations from Anthropic that both companies secretly routed millions of user conversations through Claude to train their own models. If the allegations hold, it raises serious questions about data sovereignty and the ethics of AI development on a global scale.
Shifting to hardware and privacy, Germany's data protection agency is raising red flags about Meta's smart glasses. Independent testing confirms that covering the recording LED does disable capture, but regulators say the indicator light is simply too dim and too small to be noticed unless you're standing very close. European bodies are now pushing for explicit consent from anyone being recorded, not passive assumption.
And in what may be the most quietly significant AI story of the week, a team called Vambo AI has released MORENA, a one point five billion parameter language model built specifically for twelve African languages from scratch. Despite being eight times smaller than competing models from Google, Meta, and Alibaba, MORENA outperformed all twenty six models tested, largely because it was designed around how African languages actually use tokens.
That's your digest for today. Keep surfing. Tech Beat out.
