← Evidence ledger
Safety and alignment

Google DeepMind strengthens its Frontier Safety Framework

The third framework iteration added tracked capability levels intended to detect concerning risk indicators earlier.

CURRENT ASSESSMENT · REVISION 1
AWAY FROM DOOM42confidence 90/100

Why it moved the index

Earlier tracking thresholds and a more comprehensive risk framework strengthen the chance of detecting dangerous capability growth before deployment.

AUDIT TRAIL

Assessment history

  1. R1
    Away 42 · confidence 90

    Initial source-backed launch assessment

    11 Aug 2026