Kleros (@Kleros_io)

🤖 What happens when you hand the same legal case to five different AIs? They disagree. at almost exactly the same rate human jurors do. That's one the findings in the new chapter @federicoast, @williamhwgeorge, and @robertgdean just published in AI and Arbitration (Wolters Kluwer, 2026), "When Decentralised Justice Meets Artificial Intelligence." 63 real Kleros disputes, judged by five frontier LLMs from: ChatGPT, Claude, Gemini, DeepSeek, Mistral. The takeaway isn't which model judges best. It's that you shouldn't trust a monolith AI. Round 2 of the experiment is already in flight: the team is re-running the test on real-world consumer cases from Argentina's Junín pilot and Lemon, where early evidence suggests the AIs and human jurors come to different conclusions on the same cases. Book details below ↓

🍿 Can you scale a movie critic? That's the question behind Kleros Foresight's first experiment. 16 movies, 1 judge: Kleros CTO @clesaege. Judge Dredd, Mamma Mia, 12 Angry Men, Barbie... all in the same pool. For each film: "If Clément watches this, what percentile score will he give it?" Slide a prediction higher or lower than the crowd. Closer to reality, you profit. Off, you lose. The twist: Clément won't watch all 16. Only 5 get evaluated. The top 3 by market estimate (the crowd literally decides what's worth watching), plus 1 random and 1 Clément's choice. The other 11 redeem at neutral. No profit, no loss. This is "distilled human judgement" in action. One person's taste is the ground truth, but invoking it (watching + rating a film) is slow and expensive. So the market predicts across all 16, only 5 get verified, and accurate predictions earn. The result: a recommendation signal that scales without the critic needing to watch everything. Movies are session 1. The same architecture works anywhere expert judgement exists but doesn't scale: property appraisals, grant allocation, content curation. Built on @SeerPM and @GnosisChain.