For more than a year, Polymarket's resolution layer - the UMA Optimistic Oracle - has been the subject of a slow-rolling credibility crisis. Two high-profile incidents drew the loudest reactions: a $7 million market on Ukraine's mineral deal that resolved YES while no deal was actually signed, and a $16 million UFO declassification market that resolved YES with no documents released. In between, traders watched smaller resolutions go sideways and asked the same question on Reddit and X: who actually decides what's true on Polymarket, and is the system still fair?

This article is the consolidated timeline of the UMA scandal as we have followed it from our editorial desk - what happened in the two flagship cases, why the system was structurally exposed to it, what Polymarket and UMA have done since, and what the rumored POLY token pivot would change. We also independently verify the concentration claims being circulated in a third-party investigation, showing what holds up on-chain and what does not. We point to the existing guides on our site that explain the underlying mechanics if you are new to UMA disputes.

TL;DR

  • Two flagship resolutions broke trader trust: the March 2025 "Ukraine agrees to Trump mineral deal before April?" $7M market resolving YES, and the December 2025 $16M "Trump declassifies UFO files in 2025" market also resolving YES.
  • Both were tied to UMA token concentration. A wallet known on-chain as BornTooLate.eth accumulated approximately 1.3 million UMA tokens before the Ukraine vote, becoming a top-five staker. Across the broader system, two whales reportedly held over 50% of active voting power, and as little as 5 million staked UMA can decide a contested vote.
  • Polymarket called the Ukraine outcome "unprecedented" but refused refunds, stating the market "resolved against our users' expectations" yet was not a "market failure". The UFO outcome drew similar reactions on social media without producing refunds either.
  • UMA has rolled out reforms. The most consequential is UMIP-189, which migrated Polymarket from the Optimistic Oracle V2 to the Managed Optimistic Oracle V2 (MOOV2). MOOV2 limits direct resolution proposers to a whitelist of 37 vetted addresses (Risk Labs, Polymarket employees, and trusted contributors).
  • Polymarket itself is hedging. A still-unannounced POLY token, the April 2026 launch of the CTF Exchange V2 + Polymarket USD (pUSD) collateral, and a research collaboration with EigenLayer all point to Polymarket eventually owning more of its truth layer rather than outsourcing it entirely to UMA.
  • For traders, the practical change today: ambiguous markets still carry resolution-risk premium; large markets near resolution should be sized down, not up; and the conventional wisdom of "once the question is set, the answer is mechanical" no longer holds for any market a whale could care about.

How UMA actually resolves a Polymarket market (quick refresher)

Most readers know the basics; if you are new, our UMA disputes guide walks through every step in detail. The short version:

  1. A market reaches its end date. Polymarket's adapter posts the question to UMA.
  2. A proposer submits an answer (e.g., YES or NO) along with a bond.
  3. If nobody disputes within the challenge window (typically two hours), the answer becomes final and the market settles.
  4. If disputed, the question goes to a token-weighted vote of UMA stakers. There is a commit-and-reveal period spanning roughly two days, after which the majority answer wins and bond slashing applies.

The system was designed to be optimistic - cheap and fast in the common case, and only escalating to expensive token voting when reality is contested. The 2025 incidents exposed the failure mode at the bottom of that escalation: when a vote does happen, it is resolved by token weight, and token weight can be bought.

March 2025: the $7M Ukraine mineral deal vote

The market was titled "Ukraine agrees to Trump mineral deal before April?" and asked whether Ukraine would publicly agree to a U.S. deal involving Ukrainian rare-earth mineral access by the end of March 2025.

Between March 24 and 25, 2025, the YES price moved from roughly 9% to 100% as a wave of buying overwhelmed the order book. Total notional volume in the market by resolution was approximately $7 million, large enough that the dispute window attracted the attention of researchers and reporters tracking on-chain governance activity.

On-chain analysis traced much of the resolution-relevant voting power to a wallet labeled BornTooLate.eth, which had accumulated approximately 1.3 million UMA tokens in the days before the vote. That made the wallet a top-five UMA staker, with sufficient voting weight - in combination with other concentrated wallets - to swing the resolution outcome. A separate analysis surfaced on X by the threat researcher Vladimir S. claimed a single actor with three accounts holding five million UMA tokens cast roughly 25% of the votes that resolved the market YES.

The factual question is what made this contested in the first place: no public Ukraine-U.S. mineral agreement had actually been signed by the deadline. Reporting at the time confirmed the deal was "in the works", but no signed agreement existed. Despite that, the market resolved YES.

Polymarket's response in March 2025

Polymarket posted on its Discord and on X that it had observed the manipulation. Choice quotes from the period:

  • The team called the incident an "unprecedented" governance attack.
  • The team acknowledged that the market "resolved against our users' expectations and our clarification".
  • The team stated that no refunds would be issued because, in their framing, this "wasn't a market failure" - the resolution mechanism worked as designed; the issue was that a sufficiently large UMA holder had used that mechanism in bad faith.

The reaction on X and Reddit was sharp. Many traders saw the no-refund stance as Polymarket disclaiming responsibility for an outcome users could not have priced - and the term "proof-of-whales" started circulating as a sarcastic shorthand for token-weighted governance applied to factual questions.

December 2025: the $16M UFO declassification vote

The next high-profile incident hit at the end of 2025. The market was "Trump declassifies UFO files in 2025", with approximately $16 million in cumulative volume.

UMA's on-chain vote finalized YES at 00:27:58 UTC on December 10, 2025. Per CryptoSlate's post-mortem and our own check of public records, no White House declassification bulletin matching the market's criteria had been issued: the National Archives' UAP page and Department of Defense sites showed no presidential declassification order from December 2025. The only relevant releases CryptoSlate could find were Pentagon AARO imagery from 2022 - clearly not what the market was asking about.

Two patterns from the on-chain data drew particular attention:

  • Late-session price action: large buys near 99-99.9 cents in the roughly 10 hours before resolution. Traders with no privileged information would not normally pay 99.9 cents on a contested market that close to resolution unless they had high confidence in the outcome.
  • Voter overlap with positions: post-mortem analysis noted that several large UMA token holders relevant to the resolution also held meaningful positions in the prediction market itself - the textbook conflict-of-interest that makes a token-weighted oracle dangerous on factual questions.

Community reaction this time was even sharper than in March. The phrase that circulated most widely was one user's distinction: "price manipulation is part of the game, but manipulating results through governance is not". The line captured why the UFO outcome stung more than ordinary trader losses - it was framed as a process failure, not a market failure.

Why the same failure kept happening

Both incidents share a single structural cause that reform efforts have since targeted: UMA dispute resolution is decided by token weight, and UMA token holdings are heavily concentrated.

Three numbers drive the math:

  • Of approximately 20 million UMA tokens that were stakeable around the relevant period, an individual could control as much as 7.5 million on their own.
  • Two whales reportedly controlled over 50% of active voting power between them.
  • Roughly 5 million staked UMA was sufficient to validate a vote outcome - meaning a coordinated minority with that much stake could carry a contested resolution.

Those are not the kind of numbers a system can absorb forever. The protocol's economic-security argument has always been that misbehaving voters can be slashed. But slashing only deters someone whose value-at-stake is greater than their value-from-attacking. When a $16M market resolves on a vote whose attacker holds millions in UMA - a fraction of which they might be willing to lose - the security model inverts.

Beyond concentration, two secondary factors made the failures sticky:

  • Ambiguous market wording. Both the Ukraine and UFO markets had resolution criteria that left room for interpretation. "Agrees" can mean a public verbal agreement, a draft, an in-principle handshake, a signed document - and "declassifies" is similarly fuzzy. Reform discussions consistently came back to the same point: the cleanest fix is to never let an ambiguous market reach UMA in the first place.
  • Asymmetric incentives at the proposer/voter level. If you hold both UMA tokens and a Polymarket position, every dispute is potentially a chance to vote your bag. The UMA design assumed enough independent voters to wash that out; in practice, the voter set was small enough that conflict-of-interest could compound.

The reform: MOOV2 and the whitelisted-proposer model

UMA's most consequential response to the Polymarket incidents has been UMIP-189, the governance proposal that transitioned Polymarket's resolution from the standard Optimistic Oracle V2 (OOV2) to the Managed Optimistic Oracle V2 (MOOV2).

The headline change in MOOV2 is that only whitelisted proposers can submit market resolutions directly. The whitelist comprises 37 vetted addresses - including Risk Labs (the team that maintains UMA), Polymarket employees, and trusted external contributors. Submissions from outside the whitelist are not auto-rejected, but the optimistic path - the cheap, fast resolution route - is gated to vetted proposers. The dispute path remains open if a whitelisted answer is contested.

The intended effect is twofold:

  • Cleaner proposer answers. Vetted proposers are less likely to file an opportunistic answer that benefits a single position. Anything they do propose carries a track record.
  • Higher bar for governance attacks. A bad actor cannot simply propose the answer they want and force the dispute path - they would need to propose, get rejected by the whitelist filter, and still buy enough vote weight to overturn an unaligned proposed answer. The asymmetry now favors the defender.

The trade-off is centralization. MOOV2 is by design less permissionless than OOV2. The argument from UMA's side is that the prediction-market use case was never well-served by full permissionlessness when factual claims were on the line; the argument from critics is that any whitelist creates a corruption surface of its own. Both can be true - what matters operationally is whether any of the post-MOOV2 markets see a repeat of the Ukraine or UFO patterns.

So far, no comparable resolution has surfaced under MOOV2. That is encouraging but not yet conclusive: governance attacks tend to come in clusters, and the test case will be the next $10M+ politically charged market with ambiguous wording.

Polymarket's hedge: CTF V2, pUSD, and the rumored POLY token

While UMA was pushing reform on its side, Polymarket made several moves that look, in retrospect, like preparation for owning more of its own resolution layer.

  • CTF Exchange V2 + pUSD (April 6, 2026). Polymarket announced a "full exchange upgrade" that introduced a new V2 of the Conditional Tokens Framework exchange and a native collateral token, Polymarket USD (pUSD), backed 1:1 by USDC. Practically, traders saw their balances flip from USDC.e to pUSD within a few weeks. The pUSD address is 0xC011a7E12a19f7B1f670d46F03B03f3342E82DFB on Polygon. The headline reason was unifying balances between the global and US platforms; a quieter implication is that having its own collateral token and exchange contracts gives Polymarket more design freedom around resolution mechanics it controls.
  • The unannounced POLY token. Polymarket's CMO confirmed in October 2025 that a POLY token was planned, contingent on a successful U.S. relaunch and tied to a future airdrop. The April 2026 announcement did not address POLY explicitly, but multiple reports speculate that POLY would, among other roles, take over governance and resolution functions currently outsourced to UMA - shifting the truth layer in-house.
  • EigenLayer collaboration. Polymarket and UMA jointly announced research with EigenLayer on a next-generation oracle. Whatever this produces, it is a tacit acknowledgment that the existing token-weighted UMA model is not the long-term answer for billion-dollar prediction markets.

The POLY token, if it ships with resolution authority, would be the most consequential change. Critics of UMA-as-truth-layer have argued for a year that mixing speculative token economics with factual adjudication is structurally unsound; bringing resolution in-house is one way to address that, though it raises its own questions about Polymarket's incentives as both venue and arbiter.

What this all means if you trade Polymarket today

None of the post-mortem changes the practical reality for an ordinary trader: resolution risk is still a real component of every Polymarket position, even after MOOV2. Three operating rules that we run our own trader on, after watching this story unfold:

  1. Treat ambiguity as a haircut. If the market wording leaves room for interpretation and the dollar volume is large enough to attract attention, discount your edge. The historical base rate for UMA disputes in liquid markets is meaningful, and the post-dispute outcome is binary and final.
  2. Size down at resolution, not up. The instinct on a 0.95 YES to add and lock in the last 5 cents has been wrong twice already in flagship cases. The asymmetry is brutal: small upside, large downside if a contested vote flips the answer. Our internal cap is do not buy above 0.95 on any market with even a slight resolution-risk premium.
  3. Watch the proposer/disputer pattern, not just the price. In the days before Ukraine and UFO, on-chain activity looked anomalous before the price did. Tools that surface UMA proposer/disputer behavior - including our tools and analytics overview - are now part of a sensible workflow for any market over $1M in volume.

For longer-form education on the underlying mechanics, see UMA Disputes on Polymarket and How Polymarket Markets Resolve on the guides hub. Both have been updated for the post-MOOV2 environment.

An independent investigation we tried to validate

In late April 2026 a public investigation surfaced on GitHub at github.com/sharkydelivers/uma claiming to document systemic concentration in UMA's voting layer - well beyond the single-incident framing of the Ukraine and UFO disputes. The repo includes 9 files, on-chain queries, an evidence database, and self-disclosed retractions of earlier draft claims. The author is pseudonymous, the repo has zero stars at the time of writing, and we have seen people call parts of it "lies" without offering counter-evidence. We took the investigation seriously enough to spend a few hours independently checking the strongest claims. Here is what we found - both the parts that hold up and the parts that do not.

What we independently verified on-chain

  • The Safe at 0x8180d59b...DA44A is real and substantial. Etherscan confirms it as a Gnosis Safe (Singleton 1.3.0) holding approximately $14.3 million in tokens, with the largest position being roughly 79% ACX (Across Protocol) and about 9.8% UMA. A wallet of that composition - dominated by both UMA and ACX - is by itself notable, because both protocols' voter sets have surfaced the same governance concerns over 2025-2026.
  • The Hart Lambur connection is genuine. The Safe signer at 0xcc400c09...3E37D resolves to the ENS name hal2001.eth on Etherscan, and the wallet was funded by hal2001.eth on its first incoming transaction. Hart Lambur (CEO of Risk Labs, the entity that builds UMA) used the "hal2001" handle publicly and has historically been associated with that ENS - so the address-to-person link is well-supported.
  • The DesignatedVotingV2 contract pattern is what the repo describes. The example contract at 0x0afc9a93...0B38 is in fact labeled "DesignatedVotingV2" on Etherscan, deployed on Feb 26, 2023 by an address tagged "Risk Labs: Deployer" (0x9a8f92a8...). Role-based access (the holdsRole(0) function the repo queries) is part of its public ABI.
  • Kevin Chan = maxodds.eth = the wallet from Ogle's ACX investigation. The address 0xd2a78bb8... resolves to maxodds.eth and currently holds approximately 1.92 million ACX (~39% of its portfolio), with extensive on-chain interaction with Across Protocol contracts. The link to the June 2025 Across-DAO governance scandal is independently documented by The Block, CryptoNews, CryptoPotato, and CoinDesk - that part of the repo's claims is mainstream-press-confirmed, not novel.

What we did NOT independently verify

  • "86 of 89 DV2 contracts owned by the Safe" - the repo provides reproducible queries, but exhaustively running holdsRole(0, Safe) on all 89 contracts is several hours of careful work we have not yet completed. The pattern checks out for the contract we sampled; the count claim itself is open until reproduced end-to-end.
  • "11.37M UMA = 64.4% of all staked UMA" - this requires summing balances across the alleged DV2 set plus delegations, against current total staked supply. We did not run the full reconciliation; the headline number is unverified by us as of publication.
  • "100% same-second voting timing" - the repo itself retracts this in its own narrative, noting that the subgraph time field reflects the price-request timestamp rather than the on-chain block time of each vote. The corrected claim is value-correlation, not block-timing.

Honest problems with the repo as a source

  • Pseudonymous author and zero community traction yet. The repo was created on April 21, 2026 with no public author identity and has received no obvious endorsement from independent crypto researchers as of writing. That does not make it wrong; it does mean the conclusions need independent reproduction before they belong in mainstream coverage.
  • Self-retracted earlier claims. The repo openly walks back several initial framings - including a "shared aggregator" theory that turned out to identify Coinbase 10, and a "10 contracts" undercount that became 89. The transparency is good methodologically; it also means the first draft was wrong in non-trivial ways.
  • Concentration is not the same as malfeasance. A Risk Labs-aligned Safe owning many DV2 contracts could reflect (a) consolidated voter infrastructure operated honestly, (b) coordinated decision-making, or (c) a single automated bot with multiple addresses. The on-chain data alone cannot distinguish these. The repo acknowledges this in a caveat about UMA's slashing mechanism creating rational herding incentives - rational voters lose money for dissent, so high agreement can emerge without coordination.
  • Risk Labs has publicly responded to similar accusations. When Ogle made the ACX-scandal accusations in June 2025, Hart Lambur responded that Risk Labs is a nonprofit Cayman foundation (not a private for-profit), that team members are allowed to buy ACX tokens with personal funds and vote in DAO proposals, and that maxodds.eth was publicly linked to Kevin Chan rather than secret. He called the accusations "completely untrue". That response is on the public record and any honest investigation has to grapple with it.
  • The investigation conflates structural and intentional concentration. Even if the on-chain pattern is exactly as claimed, that is consistent with both (a) a small, coordinated, profit-motivated cabal, and (b) a small protocol team that built voter infrastructure when no one else was running validators and never centralized governance enough afterward. Trader losses look the same in both worlds; the moral framing is very different.

How to verify any claim like this yourself

If you want to validate concentration claims independently - either the ones in this repo or any future investigation - the toolkit is the same:

  1. Etherscan "Read as Proxy" tab: every Gnosis Safe and every DesignatedVotingV2 contract exposes its access-control state via a public read interface. Owners and roles can be queried without a wallet.
  2. UMA voting subgraph: https://api.goldsky.com/api/public/project_clus2fndawbcc01w31192938i/subgraphs/mainnet-voting-v2/ - GraphQL endpoint where you can query revealedVotes, commitedVotes, and withdrawnRewards per voter and per round.
  3. Address labels: Etherscan's public labels ("Risk Labs: Deployer", "Coinbase 10", etc.) are crowdsourced but reasonably reliable; cross-check with two more block explorers (Blockscout, Phalcon) for any claim that hinges on a label.
  4. ENS resolution: a wallet's stated ENS can be checked at app.ens.domains independently of any third-party investigation.
  5. Funding-trace: "Funded By" on Etherscan reveals the first incoming transaction; chasing this back is the standard way to link wallets to known clusters.

What none of these tools will tell you is intent. A wallet pattern that looks like coordination on-chain can be identical to honest team infrastructure; the moral framing requires off-chain evidence (private messages, statements, court filings) the on-chain data does not contain. Treat any investigation that confidently asserts intent without that evidence with appropriate skepticism.

Where we land editorially

The structural concerns the repo raises - concentrated voter infrastructure, the same actors recurring across UMA and Across, conflict-of-interest scenarios where voters hold positions in the markets they resolve - are real and worth covering. Some of those concerns map directly to incidents that mainstream crypto press has already documented (Ogle's ACX investigation, the Ukraine and UFO outcomes on Polymarket). The repo's contribution is connecting them as a single system rather than treating each as an isolated event.

Where we do not follow the repo is in its strongest framings - that the pattern proves intentional malfeasance by named individuals. That requires evidence the on-chain data does not contain. Until either Risk Labs responds substantively to the specific 86-of-89 claim, or independent researchers reproduce the full set, we treat the repo as a useful starting point for verification work and not as a final account.

Practically for traders: nothing about this changes the playbook from earlier in this article - assume resolution risk is real, size down at 0.95+, watch UMA proposer/disputer activity on big markets. The on-chain concentration evidence, even at its weakest interpretation, supports that practical conclusion.

Sources cited

  • The Block - Polymarket says governance attack by UMA whale to hijack a bet's resolution is 'unprecedented' (Mar 2025).
  • The Defiant - Polymarket's $7M Ukraine Mineral Deal Debacle Traced to Oracle Whale (Mar 2025).
  • Yahoo Finance / Coindesk - Polymarket Suffers UMA Governance Attack After Rogue Actor Becomes Top-5 Token Staker (Mar 2025).
  • CoinMarketCap Academy - Polymarket Reports 'Unprecedented' Governance Attack by UMA Whale on Bet Resolution.
  • CryptoSlate - Polymarket faces major credibility crisis after whales forced a "YES" UFO vote without evidence (Dec 2025).
  • Webopedia - Why Is Polymarket's UMA Controversial?.
  • The Block - UMA's oracle update to limit Polymarket resolution proposals to whitelisted parties to improve market outcomes (UMIP-189 / MOOV2).
  • CoinDesk - Asia Morning Briefing: Polymarket's POLY Could Bring Oracle's Home (Oct 2025).
  • CoinDesk - Polymarket reveals a 'full exchange upgrade' (Apr 2026, CTF V2 + pUSD).
  • Polymarket Help Center - How Are Markets Disputed?; UMA Documentation - FAQs and Managed Proposers.
  • The Block - ACX drops over 10% as Across Protocol team refutes 'completely untrue' DAO manipulation and insider trading allegations (Jun 2025).
  • CryptoNews - Across Protocol Token Crashes 10% Today Amid $23M Team Misappropriation Allegations (Jun 2025).
  • CryptoPotato - Across Protocol Team Accused of a $23M Grab; Co-Founder Responds (Jun 2025, Hart Lambur response on the record).
  • Independent investigation - github.com/sharkydelivers/uma (Apr 2026, pseudonymous, on-chain queries documented). We independently verified the Safe wallet, Hart Lambur ENS link, DV2 contract pattern, and Kevin Chan-maxodds.eth-ACX connection. We did not independently reproduce the full 86-of-89 ownership count or the 64.4%-of-stake aggregate as of publication.
  • Etherscan - direct on-chain reads we ran during research: Safe 0x8180d59b...DA44A, signer 0xcc400c09...3E37D, voter wallet 0xd2a78bb8...999F7F, DV2 contract 0x0afc9a93...0B38.