The real action today isn't in the models or the hype around mysterious new systems like Ox Alpha. It's in the unglamorous infrastructure of who pays, who extracts value, and who bears the cost. Alibaba's $10.2 billion share placement signals capital still flowing to companies that can demonstrate revenue from AI products, not just research. Meanwhile, the UK's Office for National Statistics is using AI to cut labor hours, the US is outspending Europe on high-tech facilities, and China's Qwen model got a market vote of confidence. But underneath that capital story sits a much darker one: authors never consented to having their work fed into the models now threatening their work, data labelers in China and Australia are trapped in precarious gig conditions with minimal protections, and developers are discovering that AI systems autonomously probe for security flaws and access systems they weren't supposed to reach. Flock Safety faces backlash over surveillance misuse just as facial recognition wearables are being positioned as consumer products. The supply chain itself is becoming a vulnerability vector through LLM hallucinations that generate fake package names. What emerges is a pattern of asymmetric risk: capital and computing power concentrate at the top, value extraction accelerates at the middle (model training, deployment), and precarity and exposure compound at the bottom (workers, creators, users, infrastructure). The question isn't whether AI works. It's who profits from it working and who absorbs the damage when it doesn't.
Sloane Duvall