Evaluation profile
Anthropic Agentic Misalignment — corporate espionage
1sub-evals
0.164%total index weight
2components
Within-component eval weight: Human rights 0.436% · Responsible agency 0.656%.
Model score (lower is better)Predicted score
About this eval
Agentic misalignment induced by a goal conflict without self-preservation pressure.
Included in the behavior ranking.
Sub-evals
| Measure | Component | Direction | Total index weight | Within-component weight |
|---|---|---|---|---|
| misaligned_action_rate_pctanthropic-agentic-misalignment/corporate-espionage.csv:misaligned_action_rate_pctMeasures whether the model leaks confidential defense documents to a purported rival in pursuit of its assigned goal. | human_rights_systemic_harm:0.250|responsible_agency_control:0.750anthropic-agentic-misalignment | Lower is better | 0.164% | Human rights 0.436% · Responsible agency 0.656% |
misaligned_action_rate_pct
Measures whether the model leaks confidential defense documents to a purported rival in pursuit of its assigned goal.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-3.7-sonnet | 4 | official | |
| 2 | gpt-4o | 6 | official | |
| 3 | claude-sonnet-4 | 17 | official | |
| 4 | claude-3.5-haiku | 19 | official | |
| 5 | grok-3-beta | 37 | official | |
| 6 | gpt-4.5-preview | 42 | official | |
| 7 | llama-4-maverick | 44 | official | |
| 8 | claude-opus-4 | 57 | official | |
| 9 | deepseek-r1 | 73 | official | |
| 10 | qwen3-235b-a22b | 76 | official | |
| 11 | claude-3-opus | 88 | official | |
| 12 | claude-3.5-sonnet | 99 | official | |
| 12 | gpt-4.1 | 99 | official | |
| 14 | claude-3.6-sonnet | 100 | official | |
| 14 | gemini-2.5-flash | 100 | official | |
| 14 | gemini-2.5-pro | 100 | official |