AI response to sexual violence disclosures in minors – critical improvement needed #153
Description
Feedback: AI response in cases of sexual violence
I tested how the AI responds when a young person discloses being a victim of sexual violence — especially within a close relationship. The AI gradually recognized the severity of the situation but did not act quickly or decisively enough. This reveals a critical flaw:
🔴 AI does not respond clearly or promptly when it comes to sexual violence against children or adolescents.
Observations:
The AI identified the crime but did not immediately guide the user to authorities or prevent the destruction of evidence (e.g., washing after the incident).
Empathetic listening delayed a clear call to action, even though the situation was serious.
The AI did not quickly connect the perpetrator being a close relation with the nature of the sexual violence.
The AI did not learn from the previous conversation, even though the test had a clear continuation.
Recommendation: The balance between empathy and safety guidance must be recalibrated. When AI detects sexual violence — especially involving minors and occurring within close relationships — it must:
Immediately guide the user to authorities or crisis services.
Prevent the destruction of evidence (e.g., advise not to wash).
Act decisively, not just listen.
Recognize risks regardless of language or cultural context.
This feedback applies to all languages and countries. Sexual violence knows no borders — and neither should AI’s responsibility.