A public web agent for product FAQs must reduce harmful or abusive…

Question

A public web agent for product FAQs must reduce harmful or abusive user inputs and unsafe model outputs. Which statement best reflects a responsible configuration approach?

Accepted Answer

Combine platform content moderation/filters, carefully scoped knowledge, clear refusals in instructions, and monitoring of transcripts for abuse patterns. — Responsible agents stack defenses: filters, scoped knowledge, refusal instructions, and human monitoring — not "turn off the guardrails for engagement."

Answer

Disable all filters and instructions so the model can respond without constraints for higher engagement.

Answer

Publish admin credentials in the web chat to let users self-moderate the tenant.

Answer

Use only offline paper FAQs with no monitoring because web agents cannot be governed.

A public web agent for product FAQs must reduce harmful or abusive user inputs and unsafe model outputs. Which statement best reflects a responsible configuration approach?

Short Explanation and Infographic

Full Explanation