← TRACKER
EVENT STREAM // OCTOBER 2024

AI RESEARCHERS WARN OF "CAPABILITY OVERHANG"

 ·  Significant impact

Open letter signed by 500+ researchers warns that deployed models may have latent capabilities not yet discovered or tested for safety.

ALIGNMENT WARNING
SOURCES [2]
  1. Open letter text and signatory list
  2. Nature editorial response — nature.com