Natural-Language Policy Violations
Medium
- Confidence
- 93% confidence
- Finding
- The prompts explicitly direct the model to aggressively surface rumors, social whispers, and other unconfirmed market signals without requiring strong verification or explicit user acknowledgement of speculative content. In a financial-analysis skill, this increases the chance of generating misleading or manipulative-seeming output based on weak evidence, which can cause downstream harm if users treat the results as actionable intelligence.
