An edtech company is preparing to launch a Foundry-hosted tutoring chatbot for children under age 13. Before launch, the team wants to reduce two specific risks: the chatbot exposing kids to inappropriate content, and the chatbot confidently stating incorrect facts about the curriculum. (Select TWO.)
Select all correct answers, then click Submit.
Short Explanation
You need two fixes: strict content safety filtering blocks risky material before kids see it, and retrieval grounding ties answers to the school's vetted curriculum so it isn't guessing. Raising temperature or swapping to a cheaper smaller model doesn't solve either risk.
Full Explanation
A children's tutoring chatbot needs safeguards against two distinct failure modes: exposure to inappropriate content, and confidently wrong facts about the curriculum. Enabling content safety filtering at a strict severity threshold addresses the first risk by blocking outputs that fall into harmful categories before they reach a child. Grounding the assistant's answers in the school's vetted curriculum materials through retrieval addresses the second risk by anchoring responses to a trusted, curated source rather than the model's general training knowledge, reducing the chance of confident fabrication. Raising the temperature setting to make answers more varied and engaging is incorrect because it increases the variety and unpredictability of wording, which tends to make hallucinated facts more likely, not less, working against the factual-accuracy goal. Switching to a smaller, cheaper model to reduce operating costs is incorrect because it is a cost decision that says nothing about content filtering or factual grounding, and a smaller model may actually be more prone to both inappropriate outputs and inaccurate answers if it is chosen for price alone.