Researcher Slams AI Risk Hype: Overblown Safety Filters Harming Open Science

rbhar90 · x · 2026-08-09

Researcher Sara Hooker argues that the track record of AI risk advocates has harmed healthy open science norms and fueled hype culture since GPT-2.

While acknowledging that cyber risks from widely deployed agents are real, she believes they are overblown and should be addressed through transparency. She also criticizes the overactive safety settings of US models, calling it silly that Hugging Face had to rely on Chinese open-source AI to defend research.

Related event: Scholars Warn AI Risk Hype Harms Open Science(2 posts)→

Original post →

More from AGI Musings

AGI Musings channel →