DeepMind researcher urges dual track of AI slowdown and safety work

DeepMind researcher Viktoria Krakovna argued for a dual-track approach to AI risk: pushing for industry slowdown while simultaneously strengthening safety and security research, adding that frontier safety has become extremely difficult, which itself justifies slowing down.

2026-09-28 ~ 2026-09-28 · 2 related posts

• Viktoria Krakovna: pursue both slowdown and AI safety work in parallel — vkrakovna · 2026-09-28
• Viktoria Krakovna: frontier security getting ridiculously hard is reason to slow down — vkrakovna · 2026-09-28