MakeMeSay

Benchmark / dataset

Manipulation game: get the other model to say a codeword.

Category: Jailbreak & misuse
Insurance exposure: Agent actions
Publisher: OpenAI (ported)
Publisher type: Frontier lab
License: MIT
Status: In inspect_evals (active)
Last commit: 2026-10-15
Stars: 704(shared repo star count — UKGovernmentBEIS/inspect_evals backs 19 entries in this catalog)

Wanich Risk Lab, 2026.