Web3 Has No Safe AI. DMind AI Just Quantified the Gap — and KDD 2026 Made It Official.
DMind AI's new benchmark reveals that no AI model is ready for Web3's high-stakes tasks.Summary
DMind AI, in collaboration with researchers from Zhejiang University and Nanyang Technological University (NTU), has announced that its research paper "DMind Benchmark: Toward a Holistic Assessment of LLM Capabilities across the Web3 Domain" has been accepted at KDD 2026. The benchmark evaluates 31 top AI models, including GPT-5, Claude, and Gemini, across 3,543 expert questions. The verdict is clear: no system is ready for the field's highest-stakes tasks, particularly in security vulnerability detection and token economics reasoning. The benchmark comprises 3,543 expert-curated questions spanning nine core Web3 domains and is designed to be contamination-aware, ensuring models cannot cheat by memorizing answers. Since its open-source release, the benchmark has gained significant traction, reaching the #1 trending position on Hugging Face. The research highlights the critical need for rigorous, domain-grounded standards in Web3 AI, where a single reasoning error can result in catastrophic financial loss.
(Source：Crypto Reporter)