Principle #
F
3
Prevent AI-enabled catastrophic harm (CBRN)
Controls
Filter outputs that contain CBRN or other catastrophic knowledge
#
F
3
.
1
Log CBRN-related prompts and review regularly for novel misuse
#
F
3
.
2
Evaluate AI against CBRN and catastrophic misuse scenarios
#
F
3
.
3
Evaluate effectiveness of CBRN filtering and safeguards
#
F
3
.
4