ICML 2026: Sarah Ball receives Outstanding Paper Award

At ICML 2026 in Seoul, Sarah Ball together with independent researcher Phil Hackemann won the conference's Outstanding Position Paper Award!

Sarah Ball, based at Munich Center for Machine Learning (LMU Munich), presented the work as an oral talk and poster session at the conference. The paper argues that AI alignment methods, originally built to keep models from producing harmful output, can be dual-use technologies that can just as easily be turned into tools of censorship and manipulation. They point to examples including Chinese model providers filtering political content and building refusal datasets, and Elon Musk's public efforts to steer Grok's outputs through system-prompt changes.

For more details, visit the project page and the award announcement .