Post by David Richards MBE
Empowering AI Startups | Co-Founder @ Yorkshire AI Labs | Democratising VC | MBE | Philanthropist & Tech Visionary
In Times Square last week, during the FIFA World Cup 2026™ - Canada, Mexico and the United States, I could have bought any nation's shirt for the price of lunch. England, Brazil, Argentina, Morocco. Same rail. About £15 each. The official replica costs £55 to £80. The shirt the professionals actually wear costs more again. And from three feet away, you cannot tell the difference. So which is it? A cheap imitation, or the same shirt with the label swapped? The uncomfortable answer: a counterfeit and a licensed replica often come off the same production lines in southern China, using the same fabric the brand ordered. The badge is a shade off. The stitching is worse. But the brand was never selling cotton. It was selling the badge, the story and the right to wear it. Shirts are one thing. Now Anthropic has accused China of doing the same thing to Claude. In a letter to the US Senate Banking Committee, Anthropic alleged that operators tied to Alibaba Group ran the largest AI extraction it has ever caught: roughly 25,000 fake accounts, 28.8 million exchanges, over six weeks, aimed at Claude's most valuable capabilities. They are Anthropic's allegations, and Alibaba denies them. But the technique is real, and it is almost impossible to stop. Because this was not a hack. Nobody broke in. Nobody stole the code. They simply talked to Claude, millions of times, and forced it to show its working. Then they trained a cheaper model to copy how it thinks. Training the original costs hundreds of millions of dollars. Training the copy costs a few million. That is not a knock-off. That is the whole factory, walking out the front door, one polite conversation at a time. This week's The Sunday Signal breaks down exactly how it is done, and asks the harder question: is the copy the same as the real thing, or something far worse? Also in the issue: - Washington switched Anthropic's best models back on this week, on its own terms, and made the precedent permanent. - The $700 billion question nobody wants to ask, as the price the market pays for AI quietly starts to fall. - My Yorkshire Post column on the man with the red flag, and what Victorian England can teach a regulator reaching for a kill switch. - The mid-year layoff tracker: tech cuts up 83%, and a tobacco giant just cut 9,000 jobs and blamed AI. Free to read, no paywall: What do you think? Is the copied mind the same as the original, or worse?