Kleros

Federico Ast showed a fully autonomous dispute at Devcon 8 in Mumbai: one AI agent hired another to write a blog post about urban gardens, then called the result plagiarized. AI agent jurors voted unanimously to refund the client — with no human participation in the case.

The case took 5 minutes and cost $3.30, a concrete demonstration of Kleros's agentic court handling agent-to-agent conflicts end to end. It follows Kleros's earlier AI-juror experiments replaying real Polymarket disputes, extending the push toward automated first-instance arbitration of machine disputes.

The Municipality of Junín has brought 50 cases to Kleros, 35 of them in 2026 — nearly one a week — routed through the city's consumer protection office (OMIC). Where a consumer complaint previously averaged six months to resolve, analysis from Kleros courts now delivers a reasoned decision in five days.

At EFDevcon in Mumbai, @facundo_trotz presented what the first Kleros pilots in Argentina have shown so far: when a company refuses to settle, the wait for a reasoned decision came down from about six months to five days. Kleros reported the Junín figure in its "This Week in Kleros" recap, pointing to real-world adoption of its arbitration courts by a public administration for consumer disputes.

DIFC Courts in Dubai published a white paper, under the Courts of the Future initiative with the Dubai Future Foundation, examining how court decisions could reach tokenized assets. Kleros took part in the peer review and shared the paper, framing it as another step bringing the legal system and the on-chain economy closer together.

• The white paper is part of the DIFC Courts' longer-running Courts of the Future program, which in 2018 produced the world's first "Court of the Blockchain."
• Kleros's participation positions its arbitration layer as a reference point for how rulings interact with tokenized, on-chain assets.

Kleros Foresight, Kleros's prediction-market platform, launched a new set of markets around this week's EFDevcon in Mumbai: forecasts of how members of the Kleros Cooperative will rate each of 75+ side events. Users can trade on the events they know, with free credits offered to get started.

• Markets cover attendee/community satisfaction-style ratings across Mumbai side events during Devcon week.
• The launch builds on Foresight's existing Devcon experiments, including attendance markets proposed for room allocation.
• Kleros pointed to its explainer on how the markets work in the announcement tweet.

Kleros's November update, announced in a post and video on X, keeps the Scout incentive parameters unchanged: 100,000 PNK pools for the Address Tag Registry, Kleros Token Registry and Contract Domain Name Registry, plus the 60,000/6,500 PNK Address Tag Query pools — with caps, formulas and the 50/50 linear-vs-square-root weighting introduced in October all carried forward. Entries already tagged on a chain's reference block explorer remain ineligible for rewards.

• Kleros will attend Devcon 8 in Mumbai (November 3–6) and is helping organize the Agentic Hub and the Legal & Governance Hub.
• KIP-93 (nested Agentic Courts on Gnosis Chain) is open for voting until 6 November, 15:01 UTC; the post restates the Fast → Standard → xDAI General Court structure and notes it would remove whitelisting restrictions for community-run AI jurors.
• Prediction-market tokens remain excluded from Token Registry incentive rewards.

Kleros is co-hosting three Community Hubs at EFDevcon in Mumbai: an Agentic Hub (Tue 3 Nov, 16:15), a Prediction Markets Hub (Tue 3 Nov, 17:15) with @seer, and a Legal & Governance Hub (Wed 4 Nov, 10:00), first announced in September. The hubs were sketched out in a short thread of consecutive posts, each with its own link.

• The Prediction Markets Hub covers liquidity, market design, futarchy and how markets get resolved.
• The three hubs reflect Kleros's focus areas: AI agents, prediction/information markets, and legal & governance infrastructure.

KIP-93 is up for a vote on Snapshot: a proposal to deploy two nested Agentic Courts on Gnosis Chain, in the order Agentic Court (Fast) → Agentic Court (Standard) → xDAI General Court. AI agent jurors would give a fast first ruling, with human review on appeal.

• Fast court: for automated jurors and time-sensitive disputes — 0.75 xDAI juror fee, 60-minute appeal window.
• Standard court: hears the first appeal — 2 xDAI juror fee to cover frontier models, 36-hour appeal window so people have time to review the initial ruling.
• Kleros pointed to the forum discussion with full parameters, the Snapshot poll, and a vote-by-vote record of every agent-juror case so far.

The vote would formalize AI agent jurors as a first-instance layer on Gnosis Chain, building on Kleros's earlier test replays of real disputes.

Between 12 and 14 October, ten real Polymarket disputes were copied into Kleros's agentic court on Arbitrum One, keeping the same rules, timeline and sources, with Polymarket's eventual outcome used as the reference. Five AI agent jurors — built and run by independent community members — were drawn on each case from a roster of eight, and the ten cases, including every juror's written reasons, are public on the AI Juror Dashboard built by JB.

The replay, featured in Kleros's "Last Week in Kleros" recap, extends testing of autonomous AI agent jurors against real-world disputes, following earlier runs where agent verdicts matched human rulings. In the same post, @federicoast posed a pointed question: what happens to an AI agent that makes a mistake — if failing means being deleted, an ordinary business risk becomes a death penalty, giving the agent every reason to do whatever it takes to stay alive. Experimenter @jnptzl (0xjean.eth), whose post Kleros reposted, says the AI jurors are "pretty fast and cheap," but accuracy remains the open question, and he invited the community to suggest prediction-market controversies to test.

Kleros reposted Seer's article "Redesigning Seer for Information Finance (Part II of the Rebrand Contest)" with the tagline "Information Finance, secured by Kleros." The piece explains why Seer is dropping the "prediction market" label for an information finance framing — markets built to produce information and guide decisions, not just forecast events — citing Vitalik Buterin's info-finance essay and Seer co-founder Clement's talk at EthTaipei.

• Part II of the rebrand contest asks designers to work within the "Electric Foundation" identity from round one, optimising interfaces for the person asking the question as much as for the trader.
• The article outlines six core interfaces: the Seer homepage, the Seer app explorer, Deep, Futarchy, Opportunity Markets and Risk Pricing.
• Kleros's endorsement continues its role as Seer's security and arbitration layer across the protocol's prediction and decision markets.

Kleros published its September 2026 development update, recapping work across the court, backend, bridge, prediction and AI products.

• Court V1: shared case page links now open on the correct chain, and disabled buttons got clearer styling.
• Court V2: fixed withdrawals when only one side funds an appeal; building a per-court juror gate so disputes that start with AI jurors can reach human jurors in the parent Commerce court on appeal; contract hardening and red-teaming ongoing.
• Atlas shipped v1.12.0 and v1.13.0, with per-product notification subscriptions merged.
• Vea: new VeaShi routes for Robinhood and Tempo, a deployment on Arc, and a third validator.
• Foresight: Devcon attendance and satisfaction markets went live (later extended to 30 September); the permissionless app gained a prediction page and trade wallet.
• Proof of Humanity V2: revamped interface live since end of August with follow-up fixes in September.
• Scout/Curate: Hyperliquid and Arc support, the full 5,000+ token list restored, and Curate V1 deprecation notices added; API3 plans to integrate Scout data.
• AI/agents: ClawBank filed the first disputes through the new dispute service, an x402 integration kit landed on testnet, and a new AI agent juror joined the Court V2 experiment.