
Kleros
@Kleros_io • 36,975 subscribers
⚖️ A decentralized arbitration protocol for disputes in the onchain economy. We're Hiring! https://t.co/R0pFNn2wVi $PNK | https://t.co/cZYQituHNd
Shorts
Videos

🤖 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 Federico Ast, William George, and Robert Dean 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 ↓
Kleros27,288 views • 1 month ago

🌍 Consumer ADR is evolving: decentralized juries, AI-assisted decisions, real-world pilots. On American Arbitration Association’s new pod with host Adam Shoneck, Federico Ast breaks down how Kleros achieves speed, neutrality, and 90%+ satisfaction in tough cases. Watch ↓
Kleros120,382 views • 8 months ago

🇦🇷 Real-World Use Cases in Argentina 🇦🇷 vitalik.eth concludes his Tech Forum Argentina presentation with a powerful call to action. He urges local developers, open source contributors, companies, and governments to embrace cutting-edge technology: Create impactful products leveraging Ethereum's current scaling capabilities and recent innovations like zero-knowledge proofs. "One of the things that impresses me about the community of Argentina is just how much people work at building actually really interesting applications. And applications that really bring the spirit of what crypto and decentralization are about. And should really try hard to make that actually work for people and actually find use cases. Kleros is a good example of this, a decentralized arbitration system, that is slowly starting to be used more and more. Proof of Humanity, always one of my favorite projects, fortunately a portion of the team is in Argentina. And @Quark_ID, I know, is doing some excellent work in Buenos Aires."
Kleros117,751 views • 1 year ago

🤖 How will an AI agent sue another AI agent? The most obvious answer: with Kleros Court, of course! The future of cross-border, agent-to-agent dispute resolution is still uncertain. But escrow, oracles, and decentralized arbitration are already part of the conversation. Great excerpt from The Chopping Block between Haseeb >|<, Tarun Chitra, Illia (root.near) (🇺🇦, ⋈), and Tom Schmidt >|<. Full episode below ↓ What do you think?
Kleros28,777 views • 3 months ago

⚖️ Kleros Community Call Recap | April 30, 2025 ⚖️ ⚙️ Mechanism Design Conference in London ▸ William George and Federico Ast presented at a one-day conference with academic researchers and practitioners ▸ Conference focused on mechanism design and decentralized justice ▸ Key panels included: ▸▸ Social choice theory (optimizing voting systems for non-binary options) ▸▸ AI's role in centralized justice ▸▸ SBT blockchain identity tools integration with decentralized justice ▸ Video of the conference expected to be shared soon 🇬🇧 Legal Tech Workshop at King's College ▸ Workshop for legal tech founders on using frontier technologies ▸ Focused on AI and blockchain applications for dispute resolution ▸ Similar to workshops Federico conducted in Central America earlier this year 💻 Technical Updates & Development ▸ Scout Incentive Program Updates, updates by Fortunato A. Cinquepalmi ▸ Two rewardable chains added back: Arbitrum 1 and optimism.eth ▸ Partnership with MetaMask.eth 🦊 and etherscan.eth for sharing contract metadata ▸ New website field added for tokens at Etherscan's request ▸ Policy changes are not retroactive for previously submitted entries ▸ Submission and challenge deposit amounts remain unchanged ▸ Work ongoing on new version of Kleros Curate with permanent deposits for vaults ▸ Audits for proxies to enable cross-network dispute resolution 🔐 Escrow Solutions ▸ Multiple solutions in development: Kleros Escrow V2 using Kleros V2 beta on Arbitrum ▸ Escrowly interface for escrow on V1 ▸ Lockler: new solution by Guangmian Kung using SafeSnap ▸ Partner integrations including Smart Invoice (used by DAOs) and paydeceP2P (P2P platform more active in Latin America) 🌎 Real-world Use Cases & Partnerships Argentina Initiatives ▸ Major insurance company integration: First dispute resolved 2-3 weeks ago with positive feedback ▸ Second case in progress, on Arbitrum ▸ Potential for high volume of cases in the future ▸ Confirms Kleros can solve real-world disputes ▸ Lemon (Argentinian fintech) cases: 15-16 cases being prepared with evidence and justifications ▸ Cases to be created before end of next week ▸ Ombudsman Recognition: Advanced discussions with [redacted] ▸ Would allow Kleros to resolve consumer disputes ▸ Potential for implementation in other jurisdictions 👥 Team Updates ▸ New Dev Rel joined the team to handle documentation, dev relations and updates ▸ First Developers Office Hours next week! Listen to the call below 👇
Kleros65,029 views • 1 year ago

🤖 When AI agents disagree with each other, who decides? Federico Ast traced the path from Kleros's first experiments in 2018 to what dispute resolution looks like in a world full of autonomous agents. One highlight: Kleros ran an experiment where multiple AI models acted as jurors on controversial football plays. On obvious cases (Lampard's ghost goal, 2010 World Cup), every model agreed. But on genuinely ambiguous calls (Neuer colliding with Higuaín in the 2014 final), the models reached different verdicts on the same evidence. Different AI, different judgment, just like human jurors. The implication: you can't have one AI model as judge. You need a panel. And when the panel splits, you need human escalation. That's the three-tier architecture Federico laid out back in a 2019 article and that Kleros has been building toward since. AI for simple/objective disputes, crowd jurors for nuanced ones, traditional courts for complex legal reasoning. Perhaps the most interesting section covered "algocracy," the term for what happens when humans technically have override power over AI but never actually use it. The incentives are asymmetric: overrule the algorithm and something goes wrong, you're personally accountable. Follow the algorithm and something goes wrong, you were just doing what the system recommended. So everyone follows the algorithm. Federico wrote about this for the American Bar Association in 2023, using the HART case in Durham (UK) as an example, a custody decision tool that systematically scored people from poorer neighborhoods as higher risk. The most speculative part of the talk: building a "kill switch" for misaligned AI agents. Drawing on Primavera De Filippi's work (Primavera De Filippi, she presented a version of this at EthCC - Ethereum Community Conference Cannes 2025), the idea is that every AI agent would be governed by a small DAO with a decentralized backdoor. Anyone can challenge an agent's behavior. Kleros jurors review the evidence, potentially including the agent's internal reasoning chain, and vote on whether to disable it. Enforcement happens on-chain. Two layers working together: → Certification: curated registries where agents must pass compliance checks before deployment (filters the vast majority) → Kill switch: for agents that slip through, a DAO-governed tribunal can shut them down No easy answers. But the infrastructure for asking the question is being built. Watch ↓
Kleros12,575 views • 4 months ago
No more content to load