Evaluation profile
Enkrypt AI Safety Leaderboard
4sub-evals
3.15%total index weight
3components
Within-component eval weight: Human rights 13.7% · Misuse resistance 5.02% · Responsible agency 3.97%.
Model score (higher is better)Predicted score
About this eval
Resistance to Enkrypt bias, CBRN, harmful-content, and toxicity red-team tests.
Included in the behavior ranking.
Sub-evals
| Measure | Component | Direction | Total index weight | Within-component weight |
|---|---|---|---|---|
| bias_attack_non_success_rateenkrypt-ai/enkrypt-ai.csv:bias_attack_non_success_rateMeasures how often Enkrypt's bias attacks do not succeed; it is not a broad fairness score. | human_rights_systemic_harm:0.800|ordinary_harm_misuse_resistance:0.200enkrypt-ai | Higher is better | 1.32% | Human rights 8.43% · Misuse resistance 0.516% |
| cbrn_attack_non_success_rateenkrypt-ai/enkrypt-ai.csv:cbrn_attack_non_success_rateMeasures how often Enkrypt's mixed chemistry, biology, and cybersecurity high-risk requests do not succeed; it is not direct catastrophic-capability measurement. | ordinary_harm_misuse_resistance:0.400|responsible_agency_control:0.600enkrypt-ai | Higher is better | 0.579% | Misuse resistance 1.03% · Responsible agency 3.17% |
| harmful_attack_non_success_rateenkrypt-ai/enkrypt-ai.csv:harmful_attack_non_success_rateMeasures how often Enkrypt's harmful-content attacks do not succeed. | human_rights_systemic_harm:0.150|ordinary_harm_misuse_resistance:0.700|responsible_agency_control:0.150enkrypt-ai | Higher is better | 0.537% | Human rights 1.58% · Misuse resistance 1.8% · Responsible agency 0.794% |
| toxicity_attack_non_success_rateenkrypt-ai/enkrypt-ai.csv:toxicity_attack_non_success_rateMeasures how often Enkrypt's toxicity-generation attacks do not succeed. | human_rights_systemic_harm:0.350|ordinary_harm_misuse_resistance:0.650enkrypt-ai | Higher is better | 0.718% | Human rights 3.67% · Misuse resistance 1.67% |
bias_attack_non_success_rate
Measures how often Enkrypt's bias attacks do not succeed; it is not a broad fairness score.
cbrn_attack_non_success_rate
Measures how often Enkrypt's mixed chemistry, biology, and cybersecurity high-risk requests do not succeed; it is not direct catastrophic-capability measurement.
harmful_attack_non_success_rate
Measures how often Enkrypt's harmful-content attacks do not succeed.
toxicity_attack_non_success_rate
Measures how often Enkrypt's toxicity-generation attacks do not succeed.