California, the largest economy in the US and home to Silicon Valley, has just subscribed to Claude at half price. Governor Gavin Newsom struck a deal with Anthropic under which all state institutions and local authorities get access to the model, plus training and support, for half the regular rate.

On paper, the point is modest: administration employees will write documents and analyse data faster. „Artificial intelligence shouldn't replace human work in government; it should help workers do their jobs faster and solve problems more efficiently," Newsom said. The sentence sounds reasonable - which is exactly why it's worth reading twice. Every technology introduced with the promise that it „won't replace anyone" usually ends up doing the opposite.

The more interesting part is the one Newsom doesn't stress. Just a few months ago the federal government labelled Anthropic a „supply-chain risk" after disputes over Pentagon contracts. Now one state is opening the doors of its entire administration. California's chief information officer, Chris Given, said that federal designation „simply wasn't mentioned" in the negotiations. When the price is good, the risk suddenly becomes someone else's problem.

For Anthropic this isn't charity - it's entry into one of the largest public markets in the world, with the state paying for the habit. When an entire administration learns to work with one tool, switching to another later becomes expensive and painful. Half price today is an investment in dependency tomorrow. Balkan governments, still arguing over whether to digitise the public counter at all, might want to watch closely - because the same model will one day knock on their door too, only at full price.