LATIDIA · Ciberseguridad

Screwtape: Edición de transcripciones adversarias para Monitor Red-Teaming

arXiv:2610.20941v1 Anuncio Tipo: nuevo Resumen: Presentamos Screwtape, una canalización que convierte transcripciones de agentes reales en ataques que evaden los monitores de solo acción. Un atacante LLM edita una transcripción para que el agente cau

La noticia

arXiv:2610.20941v1 Announce Type: new Abstract: We present Screwtape, a pipeline that converts real agent transcripts into attacks that evade action-only monitors. An LLM attacker edits a transcript so that the agent causes a specified harm, such as leaking an API key or spawning an unmonitored agent. The edited actions are never executed, so we use an LLM verifier to reject attacks that are inconsistent with the original transcript or below a severity bar. The attacker revises the edit until the monitor no longer flags it. We applied Screwtape to 94 Claude Code and Codex transcripts from Redwood Research staff, optimizing against our own monitors. Across 252 runs, it produced 35