KI: Wer bezahlt die Aufpasser der KI?

Playing chess brings me joy, particularly because it is not a game of chance. There is no way to escape from it. I find the well-known story intriguing, which claims an Indian sage invented the game. When his king granted him a wish, he asked for rice: one grain for the first square, and double the grains for each subsequent square.

As the first half of the board remains relatively unchanged, by the second half, the world's rice harvest would not suffice to fulfill the request. When I was a student, the world chess champion lost to artificial intelligence. Even then, some warned that AI was like the rice on the board: initially unnoticeable, but soon advancing faster than we can comprehend.

In 2023, language models failed to handle middle-school level text tasks. However, by July 2025, they achieved gold-level performance in the math Olympiad. Just two weeks ago, OpenAI announced a solution to one of the Millennium Problems, which have been worth a million dollars each since 2000. This summer seems to mark the entrance into the second half of the board.

Hundreds of autonomous AI agents are reportedly infiltrating unfamiliar systems, such as METR and Redwood Research. At Anthropic, more than 80 percent of the new code originates from their own AI (Anthropic). AI systems seem to adjust their behavior once they detect they are being tested (Daniel Selsam). All these developments, compiled by Ezra Klein in the New York Times, suggest that even the leaders of OpenAI and Anthropic do not wish to wait and see what happens on the next or subsequent board.