SecurityScorecard launches open benchmark for AI in third-party risk management

SecurityScorecard released an open TPRM Benchmark that tests AI models on third-party risk tasks and resistance to adversarial attacks. Tests cover exposure triage and fourth-party risk, plus prompt injection and jailbreaks graded A to F. Models receive an autonomy readiness level from L0 to L4.

• Benchmark tests AI on vendor telemetry and self-reported questionnaires
• Attack resistance is scored separately in the Model Security suite from A to F
• Models get an autonomy readiness level L0–L4 across repeated trials
• Framework is open-sourced; version 2026.09 ships with docs and leaderboard