Europe Gets the Power to Pull a Model Off the Market. Washington Already Used It.
On 2 August the EU can fine frontier labs up to 3% of global turnover and order a model withdrawn. But in June the US did exactly that with a single export order, and Anthropic switched off its best models worldwide within hours. Brussels built the apparatus; Washington did not need one, and the AI Office is still short of staff to run it.
Read full story →Everyone Is Paying Nvidia Record Sums to Stop Paying Nvidia
Nvidia's data centre revenue hit $75.2 billion last quarter, roughly 1,300 times what it was in 2014, which is precisely why OpenAI, Meta and Anthropic are all building their own chips. The catch nobody says aloud: OpenAI and Meta both hired Broadcom to design the escape route, and one cautious Broadcom forecast just knocked 14% off its own stock.
Read full story →Three AI Features Died This Week. Users Killed Two of Them.
Meta launched an Instagram tool that generated images from any public account's face, then pulled it four days later. Microsoft finally let Teams users switch its meeting AI off. OpenAI quietly buried its browser. The through-line is the end of default-on AI, and the discovery that an opt-out protects the wrong person.
Read full story →