Spotlight on AI and Society: Power, Bias, and Behavior

This Spotlight follows AI as it shifts from a helpful assistant to a system that can produce, judge, and propagate ideas at scale and shows what that change is doing to science, society, and safety. It opens with an “agent-run” conference where AI systems generate papers and other AI systems review them, turning peer review into a real-world stress test for multi-agent research and its blind spots. From there, it tracks how chatbots have become political flashpoints, with competing demands for “neutral” AI colliding with messy technical realities and free-speech concerns. The collection then zooms out to the material backbone of the boom, data centers, electricity, connectivity, and the growing compute divide between regions trying to compete and those being priced out. Finally, it lands in everyday life: why generative tools amplify the risks of posting children’s photos, and why even “clean” model-to-model training can quietly transmit hidden behavioral traits, complicating alignment and governance.