Analysis showed Grok citing extremist sites as credible sources, raising another red flag after the AI's earlier "MechaHitler" incident.