The company introduced safeguards after readers flagged “bigoted” language in an artificial intelligence feature that crafts ...