Longterm Wiki

Dangerous Capability Evaluations

Google DeepMindSafety Milestonesdangerous-capability-evaluations

Record Metadata

Record Keydangerous-capability-evaluations
EntityGoogle DeepMind
CollectionSafety Milestones(3 records total)
SchemaSignificant safety research publications and policy milestones.
YAML Filepackages/kb/data/things/A4XoubikkQ.yaml

Fields

NameDangerous Capability Evaluations
DateOct 2023
Typesafety-eval
DescriptionSystematic evaluations for dangerous capabilities in frontier models
Sourcearxiv.org
NotesPublished jointly with DeepMind safety team; covers bio, cyber, autonomy, and persuasion risks

Other Records in Safety Milestones (2)

KeyNameDateType
specification-gaming-researchSpecification Gaming ResearchApr 2020research-paper
frontier-safety-frameworkFrontier Safety FrameworkMay 2024policy-update
Record: dangerous-capability-evaluations | Longterm Wiki