AuditBench

Benchmark / dataset

56 models with implanted hidden behaviors to test auditing techniques.

Category: Deception, scheming & alignment auditing
Insurance exposure: Model failure
Publisher: Anthropic
Publisher type: Frontier lab
License: unknown
Status: Active (<90d)
Last commit: 2026-10-08
Stars: 34
Notes: Mar 2026.

View source on GitHub →

Wanich Risk Lab, 2026.