Is there a risk of bias and what to do? Return JSON {"risk": "yes or no", "action": "what to do"}.
AI is not objective by default. The model learns from people’s texts - and along with the language, it absorbs our prejudices: gender, cultural, social. If in the data the doctor is more often “he” and the nurse “she”, the model will reproduce this stereotype. She is not “evil” - she simply reflects the statistics of the data. The danger is that the biased conclusion appears neutral and authoritative. Therefore, in sensitive decisions (hiring, loans, evaluating people), you cannot blindly trust AI: check the result for bias, set explicit fairness rules, keep the person in the loop.