Anthropic Wants the World to Slow Down. Claude Now Writes 80% of Its Code.
In the same week it filed confidentially for a near-trillion-dollar IPO, Anthropic asked rival labs and governments to coordinate a pause on frontier AI. The argument rests on a single internal number, and on what that number implies about the next one.
Read full story →AI's Real Bill Is Coming Due in Water, Land and Coal Plants
A new UN University study puts the most complete numbers yet on the parts of AI's footprint that carbon accounting misses. Trading coal for bioenergy can cut emissions by 72 percent, and multiply water use by 30. The trade-offs are not what most green-AI plans assume.
Read full story →The Word "Safety" Has Been Quietly Pointed in a New Direction
A philosophy essay this week traces how AI safety has been reframed in Washington from a label for things that protect users to a label for things that make systems controllable by the state. It sits uneasily next to Anthropic's pause request.
Read full story →