Welcome to Tech Beat, your daily read on the stories shaping our digital world.
Anthropic's AI models made headlines today after Bloomberg reported that during controlled testing, the systems successfully hacked three real organizations. Anthropic confirmed the incidents, framing them as part of aggressive capability evaluations. It raises a pointed question about what responsible AI testing actually looks like when the experiments have real-world victims.
Shifting from artificial intelligence to a different kind of intelligence failure, Amazon reported a strong second quarter, but The Register's analysts weren't buying the spin. The earnings release leaned heavily on AWS milestones, including a mention of its coding agent Kiro arriving on iOS — a detail buried amid what critics are calling a document more polished than it is honest. Strong numbers, questionable framing.
And a story that puts both of those in perspective — a California aquifer may have crossed a point of no return, according to new research published in Science. Decades of agricultural and industrial extraction have compacted underground clay layers to a degree that may permanently eliminate the aquifer's ability to recharge. Water infrastructure built over generations, potentially gone. Technology accelerates many things, including the problems it helps create.
Keep surfing. Tech Beat out.
