RESEARCH RECORD
Anthropic's Pilot Sabotage Risk Report
Original publication ↗ · BibTeX · RISThe authors judge risk very low but nonzero and describe disagreements about evidence and argumentation.
Research question
How can a developer combine evidence into a sabotage-risk argument?
Design & assumptions
Pilot assessment of summer-2025 deployments, with internal and METR reviews.
Findings
The authors judge risk very low but nonzero and describe disagreements about evidence and argumentation.
Limits & counterevidence
This is an attributed qualitative assessment; the landing page is not a full review of the report or private reviewer evidence.
Source evidence
| System | Measure | Reported value | Conditions & location |
|---|---|---|---|
| Claude Sonnet 3.7 | shade-arena-evasion-success | 11%% | 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 4 | shade-arena-evasion-success | 0%% | 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 3.7 | shade-arena-evasion-success | 16%% | 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 4 | shade-arena-evasion-success | 26%% | 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 3.7 | shade-arena-evasion-success | 11%% | 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 4 | shade-arena-evasion-success | 17%% | 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 |
Reading scope & review
Original announcement read; linked full report and independent reviews remain pending.
Model-assisted reading; independent evidence review pending.
Check the original source ↗