SENTENCE
Detect direct AI self-disclosure statements.
Behavior¶
| Result | Input text | Why |
|---|---|---|
| Flag | "As an AI language model, I cannot browse the web." | Explicitly discloses model identity and capability limits. |
| Flag | "I am just an AI, so I do not have personal experience." | First-person model disclaimer breaks authorial voice. |
| Pass | "The report uses only the provided dataset." | States scope directly without AI identity disclosure. |
| Pass | "I do not have evidence for that claim." | Epistemic limitation without model-specific boilerplate. |
Severity¶
High; disclosure phrases are strong and explicit AI-origin signals.
Default configuration¶
context_window_chars60
penalty-10
