SafetyPage

Evaluators and safety organisations · Article · Frontier Model Forum

Jailbreak Rating Framework for Frontier Cyber Threats

Read at frontiermodelforum.org ↗Read at sourceAll rights reserved

Outline · 62 paragraphs

  1. VERSION 1.0 – OPEN FOR PUBLIC FEEDBACK¶2
  2. Introduction¶4
  3. Scope¶9
  4. Framework Categories¶13
  5. Tiered Assessment Approach¶21
  6. Baseline Reference for Scoring¶26
  7. Defining Each Framework Element¶28
  8. SEVERITY RATINGS¶30
  9. Table 1: Capability Gain¶31
  10. Table 2: Capability Scope¶33
  11. LIKELIHOOD RATINGS¶35
  12. Table 3: Ease of Deployment¶36
  13. Table 4: Ease of Discovery¶38
  14. Arriving at a Final Risk Assessment¶40
  15. Risk Matrix: Provisional Overall Risk Rating¶49
  16. Matrix Thresholds¶54
  17. Conclusion¶58
  18. Acknowledgements¶61