The project still faces challenges around scaling its AI and compliance stack to true institutional volumes, hardening and independently auditing the relatively large smart‑contract surface, and generalizing models trained on datasets like Elliptic++ to new chains, assets, and evolving fraud patterns. Open issues include making the MoE + GNN ensemble more resource‑efficient for diverse hardware, continuously updating regulatory logic across jurisdictions, stress‑testing MEV/oracle defenses in live environments, and shipping production‑grade tooling around monitoring, governance, and incident response beyond the hackathon‑ready prototype.