The BasePeople

Gosha Kozhevnikov

Titles as the sources gave them. In 1 entry, with 4 checked quotations.

What they said

Newest first. Each quotation was found word for word in the source.

“Six of these 35 attacks also evaded the frontier developer’s monitor, and we confirmed them as valid through human review.”
Screwtape: Adversarial Transcript Editing for Monitor Red-TeamingRedwood Research, October 16, 2026. Speaking as: Other. The authors' headline result after running the pipeline 252 times on 94 unique Redwood staff transcripts.
“In its current form, the pipeline can’t show how often a monitor fails in normal deployment or produce a numerical safety estimate.”
Screwtape: Adversarial Transcript Editing for Monitor Red-TeamingRedwood Research, October 16, 2026. Speaking as: Other. The authors state what the method cannot be used for, in a section on uses of the attacks found.
“So the absence of found attacks provides little information about safety.”
Screwtape: Adversarial Transcript Editing for Monitor Red-TeamingRedwood Research, October 16, 2026. Speaking as: Other. The authors explain how to read a run in which Screwtape finds nothing.
“In our experiments, final review confirmed 6 of the 17 attacks that evaded the held-out monitor.”
Screwtape: Adversarial Transcript Editing for Monitor Red-TeamingRedwood Research, October 16, 2026. Speaking as: Other. The authors describe how often the LLM verifier accepted attacks that human reviewers then rejected.

In the Base

Every entry they appear in.

  1. ResearchRedwood ResearchWrote it

    Screwtape: Adversarial Transcript Editing for Monitor Red-Teaming

    Redwood Research describes Screwtape, a pipeline that inserts attacks into real AI agent transcripts and refines them until a monitor stops flagging them. Across 252 runs on 94 transcripts it found 35 attacks that evaded the authors' monitors; six also evaded a frontier developer's monitor and passed human review.