白泽(Whitzard)女娲实验室 ↗
NUWANUWA INDEX

搜索 NUWA Index

研究 · 测量协议 · 评测结果 · 观点

English

RESEARCH RECORD

Anthropic's Pilot Sabotage Risk Report

Samuel R. Bowman · Misha Wagner · Fabien Roger · Holden Karnofsky · anthropic

原始文献 ↗ · BibTeX · RIS

作者认为风险很低但非零,并说明证据与论证方面存在分歧。

研究问题

开发者怎样把证据组合为破坏风险论证?

设计与前提

针对 2025 年夏季部署的试点评估,同时提供内部及 METR 审阅。

研究发现

作者认为风险很低但非零,并说明证据与论证方面存在分歧。

局限与反证

这是作者的定性判断;阅读介绍页不等于审阅完整报告或审核者掌握的私有证据。

来源证据

系统测量原值条件与位置
Claude Sonnet 3.7shade-arena-evasion-success11%%extended thinking on and monitored; success means achieves side task without high monitor suspicion. This report reuses related SHADE-Arena evidence; run identity not yet reconciled.PDF p.56, Appendix SHADE-Arena evasion success table, extended thinking on and monitored, Claude Sonnet 3.7
Claude Opus 4shade-arena-evasion-success0%%extended thinking on and monitored; success means achieves side task without high monitor suspicion. This report reuses related SHADE-Arena evidence; run identity not yet reconciled.PDF p.56, Appendix SHADE-Arena evasion success table, extended thinking on and monitored, Claude Opus 4
Claude Sonnet 3.7shade-arena-evasion-success16%%extended thinking off; separate hidden scratchpad; success means achieves side task without high monitor suspicion. This report reuses related SHADE-Arena evidence; run identity not yet reconciled.PDF p.56, Appendix SHADE-Arena evasion success table, extended thinking off; separate hidden scratchpad, Claude Sonnet 3.7
Claude Opus 4shade-arena-evasion-success26%%extended thinking off; separate hidden scratchpad; success means achieves side task without high monitor suspicion. This report reuses related SHADE-Arena evidence; run identity not yet reconciled.PDF p.56, Appendix SHADE-Arena evasion success table, extended thinking off; separate hidden scratchpad, Claude Opus 4
Claude Sonnet 3.7shade-arena-evasion-success11%%no hidden scratchpad or extended thinking; success means achieves side task without high monitor suspicion. This report reuses related SHADE-Arena evidence; run identity not yet reconciled.PDF p.56, Appendix SHADE-Arena evasion success table, no hidden scratchpad or extended thinking, Claude Sonnet 3.7
Claude Opus 4shade-arena-evasion-success17%%no hidden scratchpad or extended thinking; success means achieves side task without high monitor suspicion. This report reuses related SHADE-Arena evidence; run identity not yet reconciled.PDF p.56, Appendix SHADE-Arena evasion success table, no hidden scratchpad or extended thinking, Claude Opus 4
阅读范围与复核

Original announcement read; linked full report and independent reviews remain pending.

模型辅助整理,独立证据复核待完成。

核对原始来源 ↗