We study a model where a large number of citizens learn a hidden state individually on an online platform. The platform receives news reports about the state and imperfectly filters misinformation in the reports, triggering conflicts about the value of the state among the citizens. We show that a platform that faces an ethical concern to internalize conflict costs due to misinformation could perversely aggravate conflicts. This cautionary observation highlights that societal efforts to mitigate conflicts, such as investments in ethical algorithm, public awareness campaigns, and government policies, are effective if and only if their implementations are sufficiently aggressive.