← All Reports
Updated report

Paxos USDG (Global Dollar)

2.2
USDG (Global Dollar)/Ethereum/Latest: August 15, 2026/Original: March 20, 2026
View full report on GitHub →View dependency graph

Score Breakdown

CategoryWeightScore
Audits & Historical20%1.50
Centralization & Control30%2.70
Funds Management30%2.50
Liquidity Risk15%1.50
Operational Risk5%1.50
Final Score2.2 / 5.0
20%30%30%15%
Low Risk

Overview

USDG (Global Dollar) is a fiat-backed USD stablecoin issued by Paxos Digital Singapore Pte. Ltd. (PDS), a Major Payments Institution supervised by the Monetary Authority of Singapore (MAS). USDG maintains a 1:1 peg to the US dollar and is fully redeemable from Paxos on a one-to-one basis (1 USDG = 1 USD).

USDG's differentiating feature is its distribution partner model — ecosystem partners (Kraken, Robinhood, Anchorage Digital, Galaxy Digital, Bullish, BitGo, KuCoin, and others) share in the yield generated by USDG reserves. This incentivizes partners to integrate USDG into their platforms, driving adoption through aligned economics rather than subsidies.

Reserves consist of cash and cash equivalents (primarily short-duration U.S. Treasury Bills) held in segregated accounts at regulated custodians, with monthly attestation reports from independent accounting firms published on the Paxos transparency portal.

USDG has material supply on 6 chains: X Layer (56.6% of supply), Solana (18.3%), Ethereum (12.9%), Robinhood Chain (10.3%), Ink (1.8%), and Hyperliquid L1 (<0.1%). A seventh deployment on Arbitrum is live but negligible (~601,207 USDG). Cross-chain bridging is handled via LayerZero V2 OFT — the mainnet OFTWrapper has configured peers for Solana, X Layer, Ink, Arbitrum, and Robinhood Chain (verified onchain, see Multi-Chain Deployments). The Hyperliquid L1 balance is a separate USDG0 HIP-1 spot token that is not a peer of the Paxos wrapper.

In February 2026 Paxos shipped a V3 upgrade that added an onchain claimable-rewards system to the token via five diamond-style facets. This is live and material: 28 payout groups with ~358.5M USDG of registered balance (~80% of Ethereum supply) are configured today. See Claimable Rewards System (V3).

Key metrics (August 15, 2026):

  • Total Supply (Ethereum): 448,595,698.67 USDG onchain. DeFiLlama's same-day snapshot reports 451.8M minted and 387.3M circulating on Ethereum; the ~64.5M reduction is DeFiLlama subtracting the Ink (62.7M) and Hyperliquid (1.8M) balances as "bridged from Ethereum"
  • Total Supply (All Chains): $3.468B — sum of totalSupply() read directly on all six EVM/Solana deployments (see Multi-Chain Deployments). This matches CoinGecko market cap ($3,467,403,004) to within 0.01%. DeFiLlama reports $3.405B because of the bridged-supply subtraction described above
  • Market Cap: ~$3.47B
  • 30-Day Supply Change: +521.3M USDG (+18.1%) — from 2,883.3M USDG (July 16) to 3,404.5M USDG (August 15) on DeFiLlama's totalCirculating.peggedUSD series DeFiLlama. The series is not monotonic: it peaked at 3,228.0M on May 15, fell to 2,889.7M by July 15, then recovered
  • DEX Liquidity (Ethereum): ~$34.3M across 8 pools, $30.5M of it the Curve USDG/USDC pool DeFiLlama
  • 24h Volume (all venues): $194.6M CoinGecko — dominated by Robinhood Chain DEXes ($130.7M); CEX venues account for ~$11.3M
  • CEX Listings: OKX, Bullish, Biconomy, Kraken, Gate (KuCoin no longer shows a USDG ticker on CoinGecko)
  • Price: $0.999877. Over the trailing 365 days the daily close never went below $0.99943 and never traded under $0.995 CoinGecko

Links:

Risk Summary

Key Strengths

  • Regulated issuer with stablecoin track record: USDG is issued by Paxos Digital Singapore under MAS supervision; the group's US entity converted to an OCC-supervised national trust bank in December 2025. Paxos has operated USDP since 2018 and PYUSD since 2023 with zero incidents across all stablecoins
  • Highest-quality reserves, now attested by KPMG: Cash and cash equivalents (primarily U.S. Treasury Bills) in segregated accounts, with monthly examinations by KPMG LLP under ISCA standards since February 2026
  • Solid audit coverage plus a live $1M bug bounty: 6 USDG-relevant audits from 3 reputable firms including Trail of Bits and Zellic, and a $1M Cantina bug bounty live since March 2026 covering the USDG contracts and cross-chain infrastructure. Source code is open (MIT license)
  • 24-hour timelock, actually exercised: Both February 2026 upgrades were routed through the TimelockController rather than executed directly — the delay is real, not nominal
  • Supply controller destination whitelisting: The two EOA supply controllers can only mint to one whitelisted Paxos address each and can only burn their own balance. A key compromise cannot mint to an attacker or seize a holder's tokens
  • Deep, verified exit liquidity: A $10M USDG→USDC swap on Curve executes at 0.023% slippage (measured onchain); depth holds to ~$15M
  • Significant market adoption: $3.47B total supply with major partners (Kraken, Robinhood, Galaxy Digital, BitGo). Strong growth trajectory

Key Risks

  • Governance consolidated into an MPC wallet — the MPC wallet (0x3Af3e85f4f97De7AD0f000B724Fb77fE5ffc024B) holds PAUSE_ROLE, ASSET_PROTECTION_ROLE, timelock PROPOSER/EXECUTOR/CANCELLER, SupplyControl SCM, four rewards roles, and OFT wrapper ownership. The MPC structure (likely Fordefi) means the key is sharded across multiple parties, but the internal quorum and policy configuration are not publicly verifiable
  • Bridge configuration is not behind timelock, upstream of a 200M USDG/24h mint path — the MPC wallet that owns the wrapper can add peers and rewrite the DVN quorum with immediate effect. This is the single largest residual risk in the system: it is the only path where one key compromise produces unbacked canonical supply with no delay and no independent circuit breaker, and unlike freeze or pause it is not required by any regulator
  • The timelock has no independent canceller — the MPC wallet is the sole holder of PROPOSER, EXECUTOR, and CANCELLER (four grants, zero revocations since January 2026). The 24h delay is a monitoring and exit window, not a circuit breaker: a compromised key can schedule a malicious upgrade and is also the only address able to cancel it
  • New rewards subsystem with hot-wallet EOAs — the V3 upgrade added 118 routed selectors, 6 roles, and three single-key EOA role-holders. It was audited by Zellic, but on a 1.2 person-week engagement, against a private repo (so deployed bytecode cannot be matched to the audited commit), and a corrective redeployment was needed two days after launch
  • Offchain reserves — reserves are entirely offchain with monthly attestation. No onchain Proof of Reserves mechanism for real-time verification
  • CEX depth is thinner than headline volume suggests — ~$11.3M across named CEX venues, with KuCoin no longer listing USDG. The $194.6M daily figure is mostly Robinhood Chain DEX volume, behind the bridge from an Ethereum holder's view
  • Relatively new (22 months) — younger than USDC (2018) or USDT (2014), though longer than many DeFi stablecoins
  • Published documentation contradicts onchain state — every role holder in the GitHub README is wrong and has been for over a year, and the README claims a multisig quorum protection that no longer exists. Neither the governance restructure nor the V3 rewards system is documented anywhere public. An integrator trusting the docs would model the system incorrectly

Critical Risks

  • Freeze/wipe capability — ASSET_PROTECTION_ROLE (held by MPC wallet) can freeze any address and wipe frozen balances. This is standard for regulated stablecoins. The MPC structure provides internal governance, but from the contract's perspective this is a unilateral capability. For DeFi integrations, a frozen vault/strategy contract would lock all USDG held by that contract
  • Upgradeable proxy with facet pattern — the USDG contract can be upgraded via UUPS proxy AND can have functional behavior changed across 118 selectors via setFacet. Both are DEFAULT_ADMIN_ROLE operations and therefore go through the 24h timelock (MPC wallet as proposer/executor)
Full Report

Contract Addresses

Core Contracts (Ethereum)

Contract Address Type
USDG Token (Proxy) 0xe343167631d89B6Ffc58B88d6b7fB0228795491D ERC1967 / UUPS Proxy (Solidity 0.8.9)
USDG Implementation 0xFACd5ff359adf87822374275699DD518Aaf9A65f USDG (Solidity 0.8.28)
Supply Control (Proxy) 0x9a7164112029b81c07636AB7b59fA813E0883BBF ERC1967 / UUPS Proxy
Supply Control Implementation 0x9e12c058a20c5b0eebaa00e44a712ec54b838971 SupplyControl (Solidity 0.8.17)

All four are source-verified on Etherscan (checked via getsourcecode), so the unverified-source critical gate does not trigger.

Token Facets (live since February 28, 2026)

The USDG proxy routes 118 function selectors to five external facet contracts via setFacet / getFacet. Facet addresses were enumerated from the complete FacetUpdate(bytes4,address) log history on the token (476 events, two deployment batches). All five are source-verified, Solidity 0.8.28.

Facet Address Purpose
PayoutGroupFacet 0x77fe0365db15a1ec05833b7c1802d68356a363df Payout-group lifecycle, address registration, group config
MultiplierMgmtFacet 0x018a090fb8d50376bf8464922ffda935a1fe2859 Reward multiplier (APR) creation and scheduled rate changes
TokenExtensionsFacet 0x7569ae2be6ac51ed7f92f8167090b52615f86497 EIP-2612/3009/1271 permit handling, cancelPermits
TokenAdminFacet 0xe90c0e2fecd06c5875b50f6b13f3b8dbdb4ce946 Admin surface (claim source, rate bounds, role plumbing)
ClaimableRewardsFacet 0x2fe2faef627a56e811e09b4e1b810b59f38f3a81 claimAll, claimForAddresses, and the …To(destination) admin variants

Governance Contracts

Contract Address Type
Token Admin (TimelockController) 0x9036566eAa5F83E0b9E1161C6c602b0Adf997654 OpenZeppelin TimelockController — 24-hour minimum delay
DEFAULT_ADMIN Multisig 0x137Dcd97872dE27a4d3bf36A4643c5e18FA40713 SimpleMultiSig — 20 owners, threshold 3
Operational Multisig (no current roles) 0x0644Bd0248d5F89e4F6E845a91D15c23591e5D33 SimpleMultiSig — 20 owners, threshold 3; no longer holds any onchain roles
Operations MPC Wallet (PAUSE / ASSET_PROTECTION / Timelock PROPOSER+EXECUTOR+CANCELLER / SupplyControl SCM) 0x3Af3e85f4f97De7AD0f000B724Fb77fE5ffc024B MPC wallet (likely Fordefi — see governance section for evidence) — holds PAUSE_ROLE, ASSET_PROTECTION_ROLE on token; PROPOSER_ROLE, EXECUTOR_ROLE, CANCELLER_ROLE on timelock; SUPPLY_CONTROLLER_MANAGER on SupplyControl

Supply Controllers

Read from SupplyControl.getAllSupplyControllerAddresses() and getSupplyControllerConfig(address) on 0x9a71…3BBF (August 15, 2026). Exactly three controllers are registered.

# Address Type Mint Limit Capacity Refill Rate Refill Window Allow Any Address Mint Destination Whitelist
SC1 0xf845a0A05Cbd91Ac15C3E59D126DE5dFbC2aAbb7 EOA 500,000,000 USDG 138,888 USDG/sec 1 hour No 0x264b…97B5 only
SC2 0x2fb074FA59c9294c71246825C1c9A0c7782d41a4 EOA 1,000,000,000 USDG 277,777.78 USDG/sec 1 hour No itself only
SC3 0x147BdE4F997f0d4C7544ED0C55eAcf1E5E6bf9c4 OFTWrapper (LayerZero bridge) 200,000,000 USDG 2,315 USDG/sec 24 hours Yes n/a (unrestricted)

Mint scope is much narrower than the capacity figures suggest. allowAnyMintAndBurnAddress is false for SC1 and SC2, and true only for SC3. Per SupplyControl.sol, canMintToAddress reverts unless the destination is in the controller's whitelist, and canBurnFromAddress reverts unless sender == burnFromAddress. A compromise of the SC1 or SC2 key therefore cannot mint to an attacker-controlled address and cannot burn any third party's balance — it can only inflate one specific Paxos-operated address and burn its own holdings. Both values were set at addSupplyController time and have never changed: zero AllowAnyMintAndBurnAddressUpdated, MintAddressAddedToWhitelist, and MintAddressRemovedFromWhitelist events over the contract's lifetime.

SC3's ceiling has been raised seven times since launch. Full LimitConfigUpdated history for the bridge controller: 10M (April 2025 launch) → 20M (Apr 22, 2025) → 30M (Sep 22, 2025) → 45M (Dec 5, 2025) → 70M (Mar 24, 2026) → 100M (May 18, 2026) → 125M (Jun 4, 2026) → 200M (Jun 30, 2026).

Note on the rate limits. Expressed as a refill window rather than a per-second rate, SC1 can mint 500M USDG per hour and SC2 1B USDG per hour. These are not meaningful constraints on Paxos itself — they are anti-fat-finger guards. The binding constraint on both EOA controllers is the destination whitelist, not the rate limit. Only SC3 has a genuinely restrictive limit (200M per 24h) relative to its unrestricted destination set.

Current balances: SC1 holds 4,493,836.57 USDG, SC2 holds 15,000,000.00 USDG, SC3 (the wrapper) holds 0 — confirming the burn-and-mint bridge model rather than lock-and-release. SC1's sole permitted mint destination 0x264b…97B5 is an EOA holding 11,622,266.31 USDG and ~5,173 ETH; it is the same address this report identifies as the MPC gas station, so it serves as both Paxos's issuance treasury and its gas-funding account.

Multi-Chain Deployments

Supply figures below are totalSupply() read directly on each chain on August 15, 2026 (all 6 decimals), not a DeFiLlama snapshot. The onchain sum ($3.468B) agrees with CoinGecko's market cap to within 0.01%; DeFiLlama's headline ($3.405B) is lower only because it treats the Ink and Hyperliquid balances as bridged-out Ethereum supply and subtracts them from the Ethereum row.

Chain Token Address totalSupply() Share LayerZero peer (eid)
X Layer 0x4ae46a509F6b1D9056937BA4500cb143933D2dc8 1,963,763,308.96 56.6% 0x9a71…3bbf (30274)
Solana 2u1tszSeqZ3qBWF3uNGPFc8TzMk2tdiwknnRMWGWjGWH 635,627,250.19 18.3% OFT store PDA (30168)
Ethereum 0xe343167631d89B6Ffc58B88d6b7fB0228795491D 448,595,698.67 12.9% — (origin chain)
Robinhood Chain 0x5fc5360D0400a0Fd4f2af552ADD042D716F1d168 356,518,732.97 10.3% 0x0d54…28d1 (30416)
Ink 0xe343167631d89B6Ffc58B88d6b7fB0228795491D 62,747,404.93 1.8% 0x9e12…8971 (30339)
Hyperliquid L1 USDG0 HIP-1 spot token (index 352) ~1,781,131 (DeFiLlama) <0.1% none — not a wrapper peer
Arbitrum 0x004b506865409877c9fa29bfb1eba929984b9bbc 601,207.02 <0.1% 0xc327…b6b6 (30110)
Total ~3,467,853,603 100%

Notes on the table:

  • Ink grew fastest over the last two weeks — 40.0M → 62.7M USDG (+57%) since the July 30 reading, while Ethereum was roughly flat (452.3M → 448.6M).
  • Arbitrum is not tracked by DeFiLlama but the deployment is live: symbol() returns USDG, totalSupply() is 601,207.02 USDG, and the mainnet wrapper holds a peer for eid 30110. Supply is negligible today, but the route is an active inbound mint path into SC3's 200M capacity.
  • Peers were enumerated onchain, not taken from documentation: all PeerSet events on the wrapper since deployment plus current peers(uint32) reads. Re-verified August 15, 2026 — still exactly five peers, no new chains added. The wrapper returns bytes32(0) for the Hyperliquid eid (30367).
  • The Robinhood Chain token address is confirmed from the wrapper itself — token() on 0x0d54…28d1 returns 0x5fc5…d168 (rpc.mainnet.chain.robinhood.com, chain id 4663).
  • Every configured inbound route requires the same 3-of-3 DVN quorum, re-verified August 15, 2026 and unchanged. Read from EndpointV2.getConfig(wrapper, receiveLib, srcEid, 2) with receiveLib = 0xc02Ab410f0734EFa3F14628780e6e695156024C2 for all five routes: required DVNs are LayerZero Labs 0x589d…236b, Canary 0xa4fe…c2cd, and Paxos 0xb0b2…2daf, with optionalDVNCount = 0. Confirmation requirements are Robinhood Chain and Arbitrum (40), X Layer and Ink (168), and Solana (32). Paxos itself operates one of the three required DVNs, so only two verifiers are independent of the issuer.
  • The wrapper's owner() and its LayerZero delegate are both the Operations MPC wallet 0x3Af3…024B, with no timelock. That address can therefore add a new peer chain or rewrite the DVN configuration unilaterally and immediately. See Governance.
  • Hyperliquid L1 is not reachable from the Paxos wrapper. The balance DeFiLlama attributes to LayerZero is a HIP-1 spot token named USDG with fullName: USDG0, tokenId: 0xae87b5246dd9f377b14bbadcf3c72131, isCanonical: false, and no linked EVM contract (Hyperliquid spotMeta API). The mainnet OFTWrapper returns bytes32(0) for the Hyperliquid eid (30367), and its full PeerSet history contains no Hyperliquid peer. The operator and control model of the USDG0 route remain unverified: this evidence establishes only that it is not a configured peer of this Paxos wrapper, not whether Paxos or a third party controls a separate adapter or custody route. Additional checks found no code at the canonical USDG address on HyperEVM and no USDG entry in LayerZero's public OFT registry. Identifying the controller requires tracing Ethereum USDG holders and Transfer events for the corresponding ~1.8M-balance escrow or recovering HyperCore deployment metadata.

Audits and Due Diligence Disclosures

Paxos has conducted 6 security audits from 3 reputable firms (Zellic, Trail of Bits, Halborn) covering the core stablecoin contracts, cross-chain integration, rewards system, and signature validation. All audits are publicly available in the paxos-token-contracts GitHub repository.

Audit History

Firm Scope Report
Zellic Core stablecoin contract review PDF
Trail of Bits Cross-chain integration PDF
Halborn Token contracts PDF
Halborn Domain separator functionality PDF
Zellic EIP-1271 signature validation PDF
Zellic USDG rewards system (V3), Feb 24, 2026 PDF

(A seventh report, PAXG V2 - Zellic, is in the same directory but scopes Pax Gold, not USDG.)

Audit firms: Zellic (3 audits) is a top-tier smart contract auditor. Trail of Bits (1 audit) is one of the most reputable security firms in the industry. Halborn (2 audits) is a well-known blockchain security firm.

Bug Bounty

Paxos launched a $1M bug bounty on Cantina on March 27, 2026, explicitly covering the USDG smart contracts and cross-chain infrastructure alongside PYUSD and PAXG, plus Web2 products, APIs, and domains. (Paxos announcement; the program is listed on the Cantina bounties page, where the max reward, live status, and March 27, 2026 start date were confirmed directly.)

Attribute Value
Platform Cantina (Cantina-triaged)
Max payout $1,000,000, paid in USDG
Status Live
Access Private / invite-only — restricted to researchers already active in the Cantina network; others must request access
Scope USDG, PYUSD, PAXG contracts and cross-chain infrastructure; public products, APIs, domains

The invite-only gating means it is not fully equivalent to an open Immunefi listing. Not listed on SEAL Safe Harbor.

Claimable Rewards System (V3)

The February 2026 V3 upgrade moved Global Dollar Network partner rewards onchain. Per the description Paxos gave Zellic, it "replaces 30-day delayed reconciliation with on-chain reward calculation" using a shares-based model with O(1) claim gas. Verified onchain state as of August 15, 2026:

  • 28 live payout groups (ids 1–30, with 20 and 29 deleted), holding a combined 358,462,181.85 USDG of registered balance — roughly 80% of Ethereum USDG supply. The largest single group (id 18) has 198,460,907.10 USDG registered.
  • Five distinct multiplier (APR) curves are in use across those groups.
  • Rewards are not minted. ClaimableRewardsBase._claimRewards debits a pre-funded claimSource address and credits the destination, reverting with InsufficientClaimSourceBalance if the source is short. The rewards system therefore adds no new mint authority — the mint-authority set remains SC1, SC2, SC3.
  • Registration does not require holder consent. registrarRegisterRewardAddress / …Batch are callable by PAYOUT_GROUP_REGISTRAR_ROLE alone. Partner-signed registration (proposeRegisterRewardAddress / acceptRegisterRewardAddress) exists but isPartnerSignedRegistrationsEnabled() currently returns false. Enrolling an address does not move its principal, but it does route the reward accrual on that balance to the group's configured destination — which may be a third party.
  • Two of the six new roles can redirect funds, per the inline documentation in Roles.sol: PAYOUT_GROUP_ADMIN_ROLE ("CAN redirect funds by changing payout destinations to arbitrary addresses") and CLAIM_ADMIN_ROLE ("CAN redirect funds by specifying arbitrary claim destinations"). Both are held by the MPC wallet. The blast radius is bounded by the claimSource balance — Paxos's own reward pool — not by user balances.

Role holders were enumerated from the token's complete RoleGranted / RoleRevoked log history and confirmed with live hasRole reads:

Role Holder Type Can redirect funds?
MULT_ADMIN_ROLE 0x3Af3…024B (MPC wallet) MPC No
MULT_RATE_ROLE 0x3Af3…024B (MPC wallet) and 0x4e43…eef8 MPC + EOA No
PAYOUT_GROUP_REGISTRAR_ROLE 0x55f7…9684 EOA No (but can enroll/unenroll any address)
PAYOUT_GROUP_ADMIN_ROLE 0x3Af3…024B (MPC wallet) MPC Yes
CLAIM_OPERATOR_ROLE 0x5fd9…2e38 EOA No (destination is group-configured)
CLAIM_ADMIN_ROLE 0x3Af3…024B (MPC wallet) MPC Yes

The three EOA holders are plain externally-owned accounts (cast code returns 0x). Roles.sol labels these role tiers "hot wallet" by design, and none of them can redirect funds or touch user principal — but they are single keys with no timelock, and this is a governance surface that did not exist at the previous assessment.

Contract Complexity

The USDG system is moderate complexity:

  • UUPS upgradeable proxy for both the token and SupplyControl contracts
  • AccessControl role-based permissions — nine distinct roles on the token: DEFAULT_ADMIN, PAUSE, ASSET_PROTECTION, plus the six V3 rewards roles. (SUPPLY_CONTROLLER_MANAGER is a SupplyControl role, not a token role — see Governance.) SupplyControl carries three more: SUPPLY_CONTROLLER_MANAGER, SUPPLY_CONTROLLER, and TOKEN_CONTRACT
  • Diamond-like facet pattern — the USDG contract uses setFacet/batchSetFacet to route 118 selectors across five external facet contracts, adding upgradeability surface area beyond the proxy. Facet changes are made by DEFAULT_ADMIN_ROLE, i.e. through the 24h timelock
  • Onchain claimable-rewards accounting (V3) — shares/multiplier model with 28 live payout groups; every transfer involving a registered address updates share accounting
  • Rate-limited minting via the SupplyControl contract with per-controller capacity, refill rates, and destination whitelists
  • EIP-2612/EIP-3009/EIP-1271 gasless and smart-wallet signature support
  • LayerZero V2 OFT bridge wrapper for cross-chain transfers
  • Freeze/wipe capability for regulatory compliance

USDG is no longer well described as "a standard ERC-20 with a pause and a freeze": the transfer path now runs through shares math for ~80% of Ethereum supply by balance.

Version History

Version Date Key Changes
v2.0.0 Nov 4, 2024 Major rewrite: consolidated Paxos stablecoins, Solidity 0.8.17, EIP-3009/2612, SupplyControl, Hardhat migration
v2.0.1 Nov 12, 2024 Bugfix: prevent frozen addresses from cross-chain transfers
v2.0.2 Aug 8, 2025 Patch: domain separator initialization fix
v2.1.0 Jan 6, 2025 EIP-1271 smart contract wallet support, dynamic DOMAIN_SEPARATOR for chain fork handling
Governance restructure ~Aug 2025–Feb 2026 Timelock deployed Jan 2026 with 24h delay; governance consolidated from multisigs to MPC wallet (Fordefi); SupplyControl admin moved from EOA to timelock; both multisigs expanded to 20 owners
v3.0 — Claimable Rewards Feb 26, 2026 Onchain rewards system: 5 facets, 118 selectors, 6 new roles. Implementation → 0x8b73…1897 (tx), executed via the timelock
v3.0.1 Feb 28, 2026 All five facets redeployed and implementation → 0xFACd…A65f (tx). Sole source diff: added name() override on PayoutGroupFacet

Upgrade history is short and fully timelocked. The token proxy has six Upgraded events in its lifetime (Oct 2024 ×1, Oct–Nov 2024 ×2, Dec 2025 ×1, Feb 2026 ×2). Both 2026 upgrades were submitted by the MPC wallet to the TimelockController (execute, selector 0x134008d3) rather than directly — confirming the 24h delay was actually exercised, not bypassed. No upgrade has occurred since February 28, 2026. SupplyControl has been on the same implementation since October 31, 2024.

Historical Track Record

  • Contract deployed: October 7, 2024 (block 20,915,336) — ~22 months in production
  • Official launch: November 1, 2024
  • Total supply: ~$3.47B across 6 chains with material supply (~$449M on Ethereum)
  • Growth trajectory: From ~$352M (mid-2025) to ~$3.47B (August 2026) — approximately 885% growth
  • 30-day change: +521.3M USDG (+18.1%)
  • Security incidents: None. No exploits, hacks, or depegging events reported
  • Peg stability: Over the trailing 365 days the daily close ranged $0.99943–$1.0351 and never traded below $0.995; the high print is a thin-liquidity artifact, not a premium regime CoinGecko
  • Paxos track record: Paxos has operated USDP (Pax Dollar, formerly PAX) since 2018 and operates PYUSD (PayPal USD) on behalf of PayPal. No Paxos-issued stablecoin has suffered a security incident or depeg

Distribution partners: Kraken, Robinhood, Anchorage Digital, Galaxy Digital, Bullish, Nuvei, BitGo, Paysafe, GSR, KuCoin, Virtual Assets Group, Tokenize

Funds Management

Accessibility

  • Minting: Available through Paxos distribution partners and direct API integration. Minting requires a Paxos account with KYC/AML verification. Not permissionless
  • Redemption: Direct 1:1 redemption through Paxos (requires account). Onchain, USDG can be exchanged via DEXes or CEXes
  • No onchain mint/redeem: Unlike USDC's permissionless onchain redemption, USDG minting and burning are controlled by Paxos supply controllers via the SupplyControl contract. End users cannot directly mint or burn
  • Fees: No fees for minting or redeeming USDG through Paxos (standard network gas fees apply)
  • Geographic restrictions: Available globally except sanctioned jurisdictions. KYC required for direct minting/redemption

Collateralization

  • Backing: 100% backed by cash and cash equivalents — primarily short-duration U.S. Treasury Bills and high-quality liquid assets held in segregated accounts at regulated custodians
  • Collateral quality: U.S. Treasury Bills are the lowest-risk financial instruments globally — backed by the full faith and credit of the U.S. government
  • Segregation: Reserve assets are held in accounts segregated from Paxos's own operating funds, providing protection in a Paxos insolvency scenario
  • Regulatory requirement: As a Major Payments Institution supervised by MAS, Paxos is required to maintain 1:1 reserves and hold them in segregated accounts
  • Offchain: All reserves are held offchain at regulated banking institutions. Token holders cannot independently verify specific reserve compositions onchain

Provability

  • Monthly attestation by KPMG: Paxos publishes monthly reserve composition reports verified by an independent accounting firm. Per the USDG Transparency page, all reports posted on or after February 27, 2026 are issued by KPMG LLP, whose examination is conducted under standards established by the Institute of Singapore Chartered Accountants (ISCA)
  • Onchain supply: Total USDG supply is verifiable onchain via totalSupply() on each chain
  • No Chainlink Proof of Reserves: No onchain oracle feed independently verifying reserves
  • Offchain verification: Reserves cannot be independently verified onchain by token holders. Must rely on the attestation reports, MAS regulatory oversight, and Paxos's institutional framework
  • Regulatory reporting: Paxos is subject to MAS supervisory requirements including regular regulatory reporting
  • MiCA compliance: USDG claims compliance with MiCA (Markets in Crypto-Assets) framework for Electronic Money Tokens under European Banking Authority oversight

Liquidity Risk

DEX Liquidity (Ethereum)

Pool data below is the DeFiLlama yields API snapshot of August 15, 2026, filtered to genuine DEX pools (lending and yield venues are broken out separately below). n/r means DeFiLlama does not report a volume figure for that pool — it is not a claim of zero volume.

Pool DEX Liquidity 7d Volume
USDG/USDC Curve $30.48M $25.06M
USDC/USDG Ekubo $1.11M n/r
syrupUSDG/USDG Uniswap V4 $0.99M n/r
USDC/USDG Uniswap V4 (2 pools) $1.17M n/r
USDC/USDG Maverick V2 $0.20M $8.38M
USDT/USDG, USDC/USDG Ekubo (2 pools) $0.37M n/r
Ethereum Total 8 pools ~$34.3M

Measured slippage on the Curve USDG/USDC pool 0xc061…1622. These are live get_dy quotes read onchain on August 15, 2026, not estimates. Pool balances are 12,362,223.19 USDG / 18,112,954.31 USDC, with A = 3000 and fee = 0.01%:

Sell size (USDG → USDC) Output Slippage
$1,000,000 1,000,007.14 −0.001%
$5,000,000 4,999,586.52 0.008%
$10,000,000 9,997,749.77 0.023%
$15,000,000 14,989,824.96 0.068%
$17,000,000 16,968,189.65 0.187%
$20,000,000 18,091,732.58 9.541%

The high amplification coefficient keeps execution nearly flat up to ~$15M, then the curve breaks hard once the USDC side of the pool (~$18.1M) is exhausted. The practical single-pool ceiling is ~$17M. The reverse direction is comparably deep ($10M USDC → USDG at 0.137%).

DEX Liquidity (Solana)

Pool DEX Liquidity 7d Volume
USDG/USDC Orca $25.69M $3.41M
USDG/USDC Kamino Liquidity $16.01M $0.22M
SOL/USDG Orca $3.51M $3.33M
USDG/ONYC Raydium $3.19M $1.47M
USDG/USX Orca $2.99M $0.67M
Other pairs Orca / Kamino / others ~$14.6M ~$11.1M
Solana DEX Total ~$66.0M

A further ~$25.7M of Solana USDG sits in lending venues (Kamino Lend, Jupiter Lend, Loopscale, Sentora), which is integration depth rather than exit liquidity.

Other Chains

Robinhood Chain has become the largest DeFi venue for USDG by a wide margin — Morpho steakUSDG ($313.5M), Morpho syrupUSDG ($85.2M), Spark Savings ($35.4M), Morpho spUSDG ($35.3M) — and carries $130.7M of the token's $194.6M daily volume across Uniswap V3/V4, Ramses, and Ekubo. This depth is real but is not reachable by an Ethereum holder without crossing the LayerZero wrapper.

Aggregate Liquidity

Source Available Notes
DEX (Ethereum) ~$34.3M 8 pools; $10M exits at 0.023% slippage on Curve alone (measured onchain)
DEX (Solana) ~$66.0M Orca and Kamino carry the depth
DEX (Robinhood Chain) high volume $130.7M/24h, but behind the bridge from an Ethereum holder's perspective
CEX OKX ($4.55M), Bullish ($4.78M), Biconomy ($1.07M), Kraken ($0.71M), Gate ($0.14M) ~$11.3M total 24h volume
Direct redemption Unlimited (via Paxos) Requires KYC account, processed during business hours
Total 24h volume, all venues $194.6M CoinGecko
  • Primary exit (permissionless): DEX swap — ~$34.3M on Ethereum, 89% of it in the Curve USDG/USDC pool. A $10M exit costs 0.023%; the single-pool ceiling is ~$17M
  • CEX depth is thin. CoinGecko shows ~$11.3M across named CEX venues, and KuCoin does not list a USDG ticker. The headline $194.6M daily figure is overwhelmingly DEX volume on Robinhood Chain, not centralized order-book depth. For an Ethereum-based integrator, CEX routing is a weaker backstop than it appears
  • Lending-market depth is not exit liquidity: a further ~$296M USDG sits in Ethereum lending/yield venues (Maple $262.0M, Aave v4 $64.7M supplied / $19.7M borrowed, Aave v3 $4.1M, Pendle ~$3.8M, Morpho PT-USDG $1.6M). Withdrawing depends on each market's utilization and does not add sell-side liquidity
  • Primary exit (KYC): Direct 1:1 redemption from Paxos — unlimited and most capital-efficient, but requires account setup and business-hours processing
  • Same-value asset: USD stablecoin — no price divergence risk from the underlying
  • No withdrawal queue: DEX/CEX exits are instant. Direct Paxos redemption follows standard processing times

Centralization & Control Risks

Governance

Token governance has been restructured from a two-tier multisig model to a model consolidated under an MPC wallet with a 24-hour timelock.

Governance was restructured between August 2025 and February 2026. The 7-owner DEFAULT_ADMIN multisig and 3-of-7 operational multisig that previously held onchain roles have been fully removed from them. Governance is now concentrated in an MPC wallet (likely Fordefi), with a 24-hour timelock on upgrades and role changes — but not on pause, freeze, or bridge configuration.

MPC wallet evidence: The operations address 0x3Af3e85f4f97De7AD0f000B724Fb77fE5ffc024B exhibits a classic MPC custody wallet pattern:

  • Gas station funding: A dedicated gas station (0x264bd8291fae1d75db2c5f573b07faa6715997b5, nonce 5.6M+, balance ~4,986 ETH, funding 62+ distinct EOAs) sends just-in-time ETH (~0.02–0.05 ETH) before each batch of operations. The account never holds large ETH balances independently.
  • Rotating gas stations: Multiple funding addresses have serviced this EOA over its lifetime (0x9195, 0x4b39, 0xf492, 0xca67, 0x264bd), consistent with MPC provider infrastructure rotation.
  • No single key holder: In MPC wallets, the private key is cryptographically sharded across multiple parties — no individual ever holds the full key. Transactions require internal policy-based approvals within the workspace.
  • Provider: The pattern (dedicated gas station with 5.6M+ nonce, just-in-time funding, multi-address servicing) is consistent with institutional MPC custody infrastructure. Paxos's website lists "Fordefi by Paxos" — a comprehensive MPC wallet platform — in its footer navigation, indicating Paxos has a direct relationship with an MPC wallet provider. This strongly suggests Paxos uses Fordefi (or similar) MPC technology for its own operational key management.

Important caveat: The internal MPC quorum/threshold and policy configuration are not publicly verifiable onchain. The security depends on the provider's implementation and Paxos's internal policy controls (e.g., requiring multiple workspace members to approve transactions). While this is significantly stronger than a single-EOA held by one person, the exact risk profile depends on the unknown internal parameters.

⚠️ Public documentation does not match onchain state. This is a standing finding, not a footnote: every role holder published in the USDG GitHub README is wrong, and the discrepancy has stood for over a year. Verified against the README and onchain state on August 15, 2026:

README states Onchain reality
DEFAULT_ADMIN_ROLE = 0x137Dcd… (multisig) TimelockController 0x903656…
PAUSE_ROLE = 0x0644Bd… (multisig) MPC wallet 0x3Af3e8…
ASSET_PROTECTION_ROLE = 0x0644Bd… (multisig) MPC wallet 0x3Af3e8…
SUPPLY_CONTROLLER_MANAGER_ROLE = 0x0644Bd… Not a token role; on SupplyControl it is the MPC wallet

The README further states that "the addresses above utilize multisignature contracts… Any change requires the presence of a quorum of signers in the same physical location." That is not merely stale — it describes a stronger and more specific protection than what exists. Governance is an MPC wallet whose quorum and policy are undisclosed, and neither of the two multisigs has held any role since August 2025.

The docs site documents no governance structure at all — no mention of the timelock, the MPC wallet, payout groups, or facets on either the overview or mainnet pages. An integrator relying on published documentation would model this system incorrectly, which is why it is treated as a live risk rather than a cosmetic gap.

All roles were re-verified onchain on August 15, 2026 via hasRole reads on the token, SupplyControl, and TimelockController, cross-checked against the complete RoleGranted/RoleRevoked log history of all three contracts. Every role assignment described below is current. No unexpected holder was found.

Tier 1 — Critical operations (upgrades, role management):

  • TimelockController (0x9036566eAa5F83E0b9E1161C6c602b0Adf997654) with 24-hour minimum delay onchain
  • Holds DEFAULT_ADMIN_ROLE and owner() on the USDG token
  • Also holds DEFAULT_ADMIN_ROLE on the SupplyControl contract (SupplyControl admin no longer an EOA)
  • Controls contract upgrades (UUPS upgradeTo), role granting/revoking, and facet changes
  • PROPOSER_ROLE, EXECUTOR_ROLE, and CANCELLER_ROLE on the timelock are all held by the MPC wallet, and by nobody else (0x3Af3e85f4f97De7AD0f000B724Fb77fE5ffc024B). The timelock's entire role history is four grants on January 22, 2026 — DEFAULT_ADMIN to itself, and proposer/executor/canceller to the MPC wallet — with zero revocations since. hasRole(CANCELLER_ROLE, …) returns false for both multisigs, the token, the issuance treasury, and the timelock itself.
  • There is therefore no independent canceller, which changes what the 24-hour delay actually buys. A timelock protects against a compromised admin only if some other party can cancel a malicious proposal during the delay. Here the sole party who can cancel is the same party who can propose. If the MPC key is compromised, the attacker schedules a malicious upgrade and is also the only address able to stop it; Paxos's recourse would be to act with the same key it has lost. The 24h window is best understood as a monitoring and exit window for integrators, not as a circuit breaker. That is a genuine protection — and it is worth materially more now that a $10M Ethereum exit clears at 0.023% slippage than it was when this report estimated exits above $5M needed CEX routing — but it is weaker than the rubric's "constrained by timelock" language implies. MPC policy controls mean multiple internal approvals are typically required to initiate transactions, which is the real mitigant here
  • The DEFAULT_ADMIN_ROLE on the timelock is held by the timelock itself (self-administered) — the timelock can grant/revoke roles on itself

Tier 2 — Operational / emergency (pause, freeze, supply management, bridge config):

  • MPC wallet (0x3Af3e85f4f97De7AD0f000B724Fb77fE5ffc024B) holds PAUSE_ROLE and ASSET_PROTECTION_ROLE directly on the token (no onchain timelock — but internal MPC policy likely requires multiple approvals) verified onchain
  • The MPC wallet also holds SUPPLY_CONTROLLER_MANAGER_ROLE on the SupplyControl contract
  • The MPC wallet is the owner() of the LayerZero OFTWrapper (0x147BdE4F997f0d4C7544ED0C55eAcf1E5E6bf9c4) and its LayerZero delegate on EndpointV2. There is no timelock on either. This is the single most consequential unmitigated power in the system: the wrapper is Supply Controller SC3 with allowAnyMintAndBurnAddress = true and 200M USDG/24h of capacity, so the same key that can call setPeer (adding a chain, or repointing an existing one) and rewrite the receive-side DVN configuration also controls a mint path to arbitrary Ethereum addresses. A compromise of this key does not require defeating the 3-of-3 DVN quorum — it can replace it
  • The former operational multisig (0x0644Bd0248d5F89e4F6E845a91D15c23591e5D33) no longer holds any roles on the token, timelock, or SupplyControl
  • SUPPLY_CONTROLLER_MANAGER_ROLE is not a token role at all. It is declared in SupplyControl.sol and appears nowhere in the token implementation or any of its five facets, whose role set is DEFAULT_ADMIN plus the eight constants in Roles.sol. Querying hasRole(SUPPLY_CONTROLLER_MANAGER_ROLE, …) on the token returns false for every address, but that is the trivial result for any role hash a contract does not use — it is not evidence of an unassigned role. On SupplyControl the role is held by the MPC wallet

Tier 3 — Rewards operations (V3, new since February 2026):

  • Six additional roles govern the claimable-rewards system. The MPC wallet holds the four privileged ones (MULT_ADMIN, MULT_RATE, PAYOUT_GROUP_ADMIN, CLAIM_ADMIN); three hot-wallet EOAs hold MULT_RATE, PAYOUT_GROUP_REGISTRAR, and CLAIM_OPERATOR. Full table and analysis in Claimable Rewards System (V3)
  • None of these roles can mint, and none can touch a holder's principal. PAYOUT_GROUP_ADMIN_ROLE and CLAIM_ADMIN_ROLE can redirect reward payouts to arbitrary destinations, bounded by the claimSource balance — a loss to Paxos and its partners, not to USDG holders
  • The registrar EOA can enrol or remove any address from a payout group without that address's consent. For a Yearn vault or strategy holding USDG this is not a fund-loss path, but it does determine who receives the reward accrual on that balance

Multisig status (neither holds active governance roles):

SupplyControl governance:

  • DEFAULT_ADMIN_ROLE on SupplyControl is now held by the Token Admin Timelock (24h delay) — this is an improvement from the previous EOA admin
  • SUPPLY_CONTROLLER_MANAGER_ROLE on SupplyControl is held by the MPC wallet 0x3Af3e85f4f97De7AD0f000B724Fb77fE5ffc024B
  • Two EOA supply controllers (SC1, SC2) have very large mint capacities ($500M/hour and $1B/hour respectively) — but both are restricted to a single whitelisted mint destination and can only burn their own balance. See the correction note under Supply Controllers. Their practical blast radius on key compromise is inflation of one Paxos-operated address, not theft from holders
  • The bridge controller SC3 is the opposite shape: unrestricted destinations, but a hard 200M USDG/24h ceiling and contract-mediated (not key-mediated) minting. Its risk is concentrated in whoever controls setPeer and the DVN config — the MPC wallet, with no timelock

Separating mandated powers from architectural ones. A regulated fiat-backed stablecoin is required to hold powers that would be red flags in a permissionless protocol, and scoring USDG down for them would double-count a trait shared by every asset in its class. It is worth stating explicitly which of USDG's admin powers fall on each side of that line, because only one side is actually informative about USDG specifically:

Power Mandated by USDG's regulatory posture? Comparable to USDC/USDT?
ASSET_PROTECTION_ROLE — freeze and wipe balances Yes — sanctions and law-enforcement compliance Yes, equivalent capability
PAUSE_ROLE — halt all transfers Yes — standard regulatory emergency control Yes
Permissioned mint/burn via SupplyControl Yes — inherent to 1:1 fiat backing with KYC issuance Yes
Upgradeable proxy Effectively yes — compliance requirements evolve Yes
Untimelocked ownership of the OFT wrapper (peers + DVN quorum) No — an architecture choice No; peers of comparable size place bridge config behind delays or multi-party control
Sole holder of proposer + executor + canceller No — an independent canceller costs nothing to add No
allowAnyMintAndBurnAddress = true on SC3 at 200M/24h No — the ceiling is a business decision Partly
Three hot-wallet EOAs on rewards roles No No
Undisclosed MPC quorum and policy No — disclosure is voluntary USDC's governance is more publicly documented

The top four rows are why USDG cannot score near the top of the governance rubric, and they are also why it should not be penalised relative to USDC — they are the price of the regulatory wrapper that makes the reserves trustworthy in the first place. The bottom five rows are the part that is specific to USDG and are not required by any regulator; they are what keep the governance score at 3.0 rather than lower.

Key governance concerns:

  1. All governance consolidated into an MPC wallet — the MPC wallet (0x3Af3e85f4f97De7AD0f000B724Fb77fE5ffc024B) holds PAUSE_ROLE, ASSET_PROTECTION_ROLE, timelock PROPOSER+EXECUTOR+CANCELLER, and SupplyControl SCM. However, as an MPC wallet (likely Fordefi), the private key is sharded across multiple parties — no single individual can unilaterally sign transactions. The risk is comparable to a multisig with an unknown internal quorum, not a standard single-key EOA
  2. 24-hour timelock is a strong improvement — the 3h delay increased to 24h, providing a meaningful monitoring window for contract upgrades. Combined with MPC policy controls, this creates defense-in-depth for critical changes
  3. Emergency actions are immediate by design — PAUSE_ROLE and ASSET_PROTECTION_ROLE are exercisable onchain without delay, which is the correct design for an emergency control; the MPC wallet's internal policy layer is what governs their use
  4. SupplyControl admin improvement — moving DEFAULT_ADMIN on SupplyControl from an EOA to the 24h timelock prevents unilateral addition of supply controllers
  5. Freeze/wipe capability — ASSET_PROTECTION_ROLE can freeze individual addresses and wipe frozen balances. This is standard for regulated stablecoins (USDC, USDT have equivalent capabilities). The MPC wallet structure provides internal governance but the onchain capability remains unilateral from the contract's perspective
  6. Internal MPC quorum unknown — while MPC is inherently multi-party, the exact number of key shards, approval threshold, and policy rules are not publicly verifiable. This is a transparency gap
  7. Bridge configuration is not behind timelock — the MPC wallet owns the OFTWrapper and is its LayerZero delegate, so peer set and DVN quorum can both be changed with immediate effect, upstream of a 200M/24h mint path. Contract upgrades get 24 hours of warning; adding a new mint route does not
  8. Rewards system introduces hot-wallet EOAs — three single-key EOAs hold operational rewards roles. They cannot mint or seize funds, so this is a bounded concern, but it reverses part of the 2026 consolidation away from EOA-held roles
  9. Supply controller destination whitelisting is a genuine mitigant — SC1 and SC2 are whitelist-bound, which is the main reason governance is not scored higher despite items 7 and 8

Programmability

  • Token: Standard ERC-20 with EIP-2612/EIP-3009 gasless support. Transfers are fully programmatic onchain
  • Minting/burning: Through SupplyControl contract with rate-limited supply controllers. Onchain but Paxos-controlled (not permissionless)
  • Exchange rate: Fixed 1:1 USD peg — no oracle needed for the token itself
  • Reserves: Entirely offchain. Reserve management, yield generation, and reporting are handled by Paxos and its custodians with no onchain visibility into specific holdings
  • Partner reward accounting is now onchain — the V3 rewards system replaced a 30-day offchain reconciliation process with onchain share/multiplier accounting settled from a funded claim source. This is a genuine programmability improvement over the model assessed in March 2026, though the reward rates (multipliers) are still set by admins rather than derived
  • Pause mechanism: PAUSE_ROLE can freeze all transfers/approvals; minting/burning remain operational during pause
  • Operations split: Token operations (transfers, approvals, reward accrual and claims) are onchain and programmatic. Reserve operations (investment, custody, yield generation) are entirely offchain and centralized

External Dependencies

  • No DeFi protocol dependencies — USDG is a standalone stablecoin, not dependent on any external DeFi protocols
  • Banking infrastructure — reserves held at regulated custodians (inherent to fiat-backed stablecoins)
  • LayerZero V2 — cross-chain bridging via OFTWrapper to Solana, X Layer, Ink, Arbitrum, and Robinhood Chain. Still non-critical for Ethereum-only usage — an Ethereum holder's balance does not depend on the bridge — but the exposure has grown on two axes since the last assessment: ~87% of total supply now sits on chains reachable through this wrapper, and the inbound receive path can mint canonical USDG on Ethereum up to 200M USDG/24h (raised from 45M in four steps between March and June 2026). The wrapper's owner/delegate is the MPC wallet with no timelock
  • Curve/Uniswap — DEX liquidity for secondary market exits (not a protocol dependency, but relevant for exit liquidity)

Operational Risk

  • Team: Paxos was founded in 2012 by Charles Cascarilla (CEO) and Rich Teo (co-founder). Paxos is a well-established, regulated fintech company with 200+ employees
  • Track record: Operates multiple stablecoins: USDP (since 2018), PYUSD (PayPal USD, since 2023), USDG (since 2024). No security incidents across any Paxos stablecoin
  • Regulation: Paxos Digital Singapore Pte. Ltd. (PDS) — the issuer of USDG — is a Major Payments Institution supervised by the Monetary Authority of Singapore. USDG also claims MiCA compliance via the EU entity
  • US entity is now OCC-supervised, not NYDFS (corrected — resolves issue #388). On December 12, 2025 the Office of the Comptroller of the Currency conditionally approved the application of Paxos Trust Company, LLC (a New York state trust company) to convert to an uninsured national trust bank, operating as Paxos Trust Company, National Association under OCC Charter Number 25379 (OCC conditional approval letter, OCC decision document, Paxos announcement). All of Paxos Trust Company's US-based activity is now subject to OCC supervision rather than NYDFS. Two qualifications matter for risk purposes: (a) the bank does not take deposits and is not FDIC-insured; (b) the conversion covers the US entity behind PYUSD and PAXG — USDG's issuer remains PDS in Singapore under MAS, so USDG's direct regulator is unchanged. The correct framing is that the Paxos group's US supervision moved from state to federal, which strengthens the group-level regulatory picture without altering USDG's own licensing
  • Documentation: Comprehensive documentation at docs.paxos.com covering integration guides, API reference, and contract addresses. Source code is MIT-licensed and publicly available on GitHub. Gaps: the GitHub README's role-holder list is stale by more than a year, and neither the docs site nor any public post documents the 2026 governance restructure or the V3 rewards system
  • Legal structure: Paxos Digital Singapore Pte. Ltd. (Singapore entity, issuer of USDG), with Paxos Trust Company, N.A. (US federal trust bank, issuer of USDP/PYUSD/PAXG) and Paxos Issuance SARL (EU entity) as sister companies

Monitoring

Key Contracts to Monitor

Contract Address Monitor
USDG Token 0xe343167631d89B6Ffc58B88d6b7fB0228795491D totalSupply(), paused(), Transfer events, Mint/Burn events
SupplyControl 0x9a7164112029b81c07636AB7b59fA813E0883BBF Supply controller additions/removals, rate limit changes
TimelockController 0x9036566eAa5F83E0b9E1161C6c602b0Adf997654 CallScheduled, CallExecuted events (24h delay — gives monitoring window)
Operational Multisig 0x0644Bd0248d5F89e4F6E845a91D15c23591e5D33 Submitted/executed transactions (currently role-less; any re-grant is notable)
Operations MPC Wallet (all governance) 0x3Af3e85f4f97De7AD0f000B724Fb77fE5ffc024B Any transactions — controls pause, freeze, timelock scheduling, supply controllers, and the OFT wrapper's peers + DVN config. MPC wallet (likely Fordefi) — key sharded across multiple parties. Monitor for unexpected transactions
OFTWrapper (SC3) 0x147BdE4F997f0d4C7544ED0C55eAcf1E5E6bf9c4 PeerSet events (new/changed chains), OwnershipTransferred, DelegateSet on EndpointV2, and receive-library/DVN config changes. Highest-leverage surface in the system that is not behind a timelock
Issuance Treasury / Gas Station 0x264bd8291fae1d75db2c5f573b07faa6715997b5 SC1's only permitted mint destination and the MPC gas funder. Monitor large USDG inflows (new issuance) and unusual ETH outflows (possible MPC key migration or provider change)

Critical Events to Monitor

  • Pause events — Paused/Unpaused on the token — all transfers stop when paused (MPC wallet can trigger immediately, but internal policy likely requires multiple approvals)
  • Freeze events — individual address freezes via ASSET_PROTECTION_ROLE — could affect DeFi integrations (MPC wallet, no onchain timelock)
  • Supply changes — large mints/burns (>5% of supply in 24h) could indicate operational issues
  • Contract upgrades — Upgraded events via UUPS proxy — 24h timelock provides advance notice via CallScheduled
  • Supply controller changes — SupplyControllerAdded/Removed via SupplyControl — SCM role held by MPC wallet, admin role under 24h timelock
  • Rate limit changes — LimitConfigUpdated. SC3's capacity has been raised seven times since launch (10M → 200M); each raise widens the bridge's inbound mint blast radius
  • Mint destination whitelist changes — MintAddressAddedToWhitelist / RemovedFromWhitelist and AllowAnyMintAndBurnAddressUpdated. Zero of these have ever fired. Any first occurrence — especially flipping allowAnyMintAndBurnAddress to true on SC1 or SC2 — would remove the principal mitigant on the EOA supply controllers and should be treated as a high-priority alert
  • OFT peer and DVN changes — PeerSet on the wrapper and receive-config changes on EndpointV2. Untimelocked, upstream of a 200M/24h mint path
  • Timelock events — CallScheduled gives 24h advance notice of all critical admin changes
  • Facet changes — FacetUpdate(bytes4,address) events indicate functional changes to the token contract without a proxy upgrade
  • Rewards system events — PayoutGroupCreated/Deleted, PayoutGroupDestinationSet, AccountRegistered/Deregistered, ClaimSourceSet, MultiplierRateScheduled. Integrators holding USDG should watch AccountRegistered for their own vault/strategy addresses, since registration is registrar-initiated and does not require consent
  • MPC wallet transactions — any transaction from 0x3Af3e85f4f97De7AD0f000B724Fb77fE5ffc024B — controls all governance actions. MPC structure provides multi-party security, but the address remains the single onchain governance point
  • Rewards hot-wallet EOAs — transactions from 0x4e43…eef8, 0x55f7…9684, 0x5fd9…2e38

Monitoring Functions

Function Contract Purpose Frequency
totalSupply() Token Supply tracking Every 6 hours
paused() Token Operational status Hourly
isFrozen(address) Token Address freeze status On integration
getSupplyControllerConfig(address) SupplyControl Capacity, refill rate, whitelist, allowAnyMintAndBurnAddress Daily
getAllSupplyControllerAddresses() SupplyControl Detect a fourth controller being registered Daily
getMinDelay() TimelockController Timelock delay changes (currently 24h) Weekly
owner() OFTWrapper Bridge admin key rotation Daily
peers(uint32) OFTWrapper New/changed bridge routes Daily
getFacet(bytes4) Token Facet routing changes On FacetUpdate
payoutGroupIdOf(address) Token Whether your own vault/strategy has been enrolled in a payout group On integration, then weekly
hasRole(CANCELLER_ROLE, …) TimelockController Whether an independent canceller has been added (improvement) or the sole-holder setup persists Weekly
Role events Token / SupplyControl / TimelockController Monitor RoleGranted/RoleRevoked — contracts use plain AccessControl, so holders cannot be enumerated and must be reconstructed from logs On change

Reassessment Triggers

  • Time-based: Reassess in 6 months (February 2027)
  • TVL-based: Reassess if total supply changes by more than ±50% from current $3.47B
  • Incident-based: Reassess after any exploit, freeze affecting DeFi protocols, depegging event, or adverse regulatory action
  • Governance-based (IMPORTANT): Reassess if governance transitions from MPC wallet back to multisig (score improvement), or if MPC wallet shows signs of compromise. Any change in role holders, timelock delay, or MPC provider/gas station infrastructure warrants reassessment
  • Independent canceller (score improvement): Reassess if CANCELLER_ROLE on the timelock is granted to any party other than the MPC wallet. This is the cheapest change Paxos could make that would meaningfully improve the governance score — it would convert the 24h delay from a monitoring window into an actual circuit breaker. Conversely, reassess if the sole-holder configuration is extended to any newly-deployed governance contract
  • MPC transparency: Reassess if Paxos discloses MPC provider, quorum, and policy configuration (potential score improvement from reduced uncertainty)
  • Governance-based: Reassess if SupplyControl admin or SCM transitions, or if multisig composition/threshold changes
  • Bridge-based (IMPORTANT): Reassess on any PeerSet event adding or repointing a chain on the OFT wrapper, any change to its owner() or LayerZero delegate, any receive-library/DVN configuration change, or any further increase to SC3's mint capacity beyond 200M USDG/24h
  • Supply-controller scope: Reassess immediately if AllowAnyMintAndBurnAddressUpdated fires for SC1 or SC2, or if a mint-destination whitelist changes. None of these events has ever occurred; the first would remove the principal mitigant on the EOA supply controllers
  • Rewards system: Reassess if the claimSource changes, if rewards roles move to or from EOAs, if partner-signed registrations are enabled, or if a Yearn-controlled address is registered into a payout group
  • Regulatory-based: Reassess if MAS takes enforcement action, if Paxos loses its MPI license, if the OCC modifies or rescinds the conditional approval of Charter 25379, or if USDG issuance moves between legal entities
  • Bug bounty: Reassess if the Cantina program opens to the public (further improvement) or is discontinued (deterioration)
  • Proof of Reserves: Reassess if onchain reserve verification (e.g., Chainlink PoR) is deployed (score improvement)
  • Attestation: Reassess if the attestor changes from KPMG LLP or the monthly cadence lapses
  • Documentation (score improvement): Reassess if Paxos brings public documentation into sync with onchain state. Operational Risk is held at 1.5 solely by the documentation column; reaching 1.0 requires (a) correcting the four role rows in the usdg-contract README, (b) removing the "quorum of signers in the same physical location" claim, and (c) documenting the governance model (timelock delay, MPC wallet, which powers are timelocked and which are immediate), the V3 rewards system, and the bridge's supply-controller role and mint ceiling

Appendix A — USDG Risk as a Held Asset in yvUSD

Context: yvUSD is a USDC-denominated Yearn V3 vault that deploys USDC into multiple yield strategies. This appendix assesses the risk of including USDG as an asset held within yvUSD strategies — e.g., lending USDC against USDG collateral in Morpho markets, or holding USDG positions as part of yield strategies.

Risk Assessment for yvUSD Exposure

Depeg Risk: LOW

  • USDG is backed 1:1 by cash and U.S. Treasury Bills at regulated custodians under MAS supervision
  • Paxos has a 7+ year stablecoin track record with zero depegging incidents across USDP, PYUSD, and USDG
  • Strong peg stability observed: price consistently at $0.999-$1.000
  • Risk is comparable to holding USDC — both are regulated fiat-backed stablecoins

Smart Contract Risk: LOW-MEDIUM

  • 6 audits from reputable firms, source code is open and verified
  • UUPS upgradeable proxy with 24h timelock + MPC wallet governance
  • Diamond-like facet pattern adds upgradeability surface area
  • Contract complexity is moderate (standard ERC-20 + AccessControl + rate-limited supply)

Freeze Risk: MEDIUM

  • Paxos can freeze any address and wipe frozen balances via ASSET_PROTECTION_ROLE (held by MPC wallet)
  • If a Yearn vault or strategy contract holding USDG gets frozen, those funds become inaccessible until unfreezing
  • This is the same risk profile as USDC (Circle can freeze addresses) — it has never been used against DeFi protocols but the capability exists
  • Mitigation: Ensure the vault/strategy addresses are known to Paxos and not on any sanctions list

Liquidity Risk for yvUSD: LOW

  • Ethereum DEX liquidity (~$34.3M) with measured execution: $10M exits at 0.023% slippage, $15M at 0.068% on the Curve USDG/USDC pool alone
  • The practical single-pool ceiling is ~$17M; beyond that, route across the other seven Ethereum pools or use Paxos redemption
  • Given yvUSD's scale, USDG liquidity is comfortably adequate — liquidity is not the binding constraint on this integration

Rewards enrolment: a yvUSD strategy address holding USDG can be registered into a Paxos payout group by the PAYOUT_GROUP_REGISTRAR_ROLE EOA without consent. This does not move or encumber the strategy's principal, but it routes the reward accrual on that balance to the group's configured destination — potentially a third party. Check payoutGroupIdOf(<strategy>) on the token at integration time and monitor AccountRegistered events for Yearn-controlled addresses.

Overall Assessment for yvUSD: USDG is a suitable stablecoin asset for yvUSD strategies. Risk is comparable to other regulated stablecoins. Position sizing should be driven by freeze risk and issuer concentration rather than by exit depth, which is ample at current scale.

Risk Factor Level Notes
Depeg Low Regulated, T-Bill backed, KPMG-attested; never below $0.995 in 365 days
Smart Contract Low-Medium 6 audits + $1M bounty, but upgradeable, and V3 added 118 facet-routed selectors on a light audit engagement
Freeze Medium Standard for regulated stablecoins, never used vs DeFi
Liquidity Low ~$34.3M DEX; $10M exits at 0.023% measured onchain
Rewards enrolment Low Non-custodial, but a registrar EOA can redirect reward accrual on a held balance
Overall Low-Medium Suitable; freeze risk, not liquidity, is the binding constraint

Appendix B — USDG Risk as Collateral for yvUSDC-1 Lending

Context: yvUSDC-1 is a USDC-denominated Yearn V3 vault that deploys USDC into lending strategies. This appendix assesses the risk of lending USDC against USDG as collateral — e.g., in Morpho markets where borrowers post USDG to borrow USDC.

Risk Assessment for Lending Against USDG

Collateral Quality: HIGH

  • USDG is 1:1 backed by U.S. Treasury Bills — same quality as USDC collateral
  • Regulatory framework (MAS supervision) provides strong assurances on reserve integrity
  • No history of depegging or reserve shortfalls
  • Appropriate for same-value lending (USDG collateral for USDC borrows)

What the major lending markets actually do with USDG

Every major Ethereum lending market assigns USDG a collateral factor of zero — it is listed as a borrowable/supply asset, not as collateral. Read onchain August 15, 2026:

Venue USDG role LTV / Collateral Factor Liquidation threshold Notes
Aave v3 (Ethereum) Borrow-only 0 0 usageAsCollateralEnabled = false, borrowingEnabled = true, active, not frozen. Borrow cap 100M, supply cap 120M, reserve factor 20%. Read via AaveProtocolDataProvider.getReserveConfigurationData on 0x0a16…bECD
Aave v4 (Core spoke 0x94e7…c485) Borrow-only collateralFactor = 0 maxLiquidationBonus 10000 (= 1.0×), liquidationFee 0 Reserve id 11. 64,683,806.35 USDG supplied, 19,662,983.50 borrowed. Same treatment as EURC, RLUSD, frxUSD, GHO in that spoke — while USDC and USDT both carry CF 7800
Morpho Blue (Ethereum) Loan asset only n/a n/a Zero markets use USDG as collateral. Three dust-sized markets (WBTC, wstETH, cbBTC collateral) borrow USDG at 86% LLTV
Maple Lending pool asset n/a n/a $262.0M syrupUSDG — the largest single USDG venue on Ethereum

The pattern is unambiguous and consistent across independent risk managers: USDG is treated as a good asset to lend and a poor asset to collateralize, most plausibly because of the freeze/wipe capability discussed below rather than because of any depeg concern. Aave's own risk process assigns USDC and USDT an 78% collateral factor in the same spoke where USDG gets zero.

Liquidation Risk: MEDIUM

  • DEX liquidation path: the Curve USDG/USDC pool ($30.5M liquidity) is the primary liquidation venue on Ethereum. Measured onchain: a $10M liquidation executes at 0.023% slippage and a $15M liquidation at 0.068%
  • Liquidation depth vs exposure: the hard constraint is the ~$18.1M USDC side of the pool. A $20M single-block liquidation would incur ~9.5% slippage; positions above ~$15M need multi-block execution or routing across the other seven Ethereum pools
  • Same-value asset: Since USDG and USDC are both USD stablecoins, liquidation is essentially a stablecoin-to-stablecoin swap — much lower risk than volatile collateral liquidations

Freeze Risk for Lending: MEDIUM-HIGH

  • If USDG collateral in a Morpho vault gets frozen by Paxos, the lending protocol cannot liquidate that collateral
  • This creates bad debt risk: the borrower defaults, but the collateral is frozen and cannot be seized/sold
  • This is the most significant risk for lending against USDG — the freeze capability creates a scenario where collateral becomes illiquid and unliquidatable
  • Mitigation: This risk exists for all regulated stablecoins (USDC, USDT) used as collateral and has never been triggered against DeFi protocols. MAS supervision provides accountability

Counterparty Risk: LOW

  • Paxos is well-regulated with zero incidents — very low probability of voluntary freeze of legitimate DeFi contracts
  • Involuntary freezes (law enforcement, sanctions) would target specific addresses, not the broad DeFi ecosystem
  • Regulatory clarity is improving, reducing the probability of blanket DeFi restrictions

Overall Assessment for yvUSDC-1: USDG is acceptable as lending collateral on its own merits — T-Bill backing, a clean 22-month peg record, and deep same-value liquidation liquidity. The parameters below are conservative because no major venue currently lends against it at all: where USDG is used in production, it is as the loan asset, not the collateral.

Risk Factor Level Notes
Collateral Quality High T-Bill backed, MAS-supervised issuer, KPMG-attested, 1:1 with USD
Liquidation Low-Medium ~$30.5M Curve depth; $10M liquidates at 0.023%, but a hard cliff past ~$17M
Freeze (Collateral) Medium-High Frozen collateral = unliquidatable = bad debt risk
Market acceptance Notable Aave v3, Aave v4, and Morpho all assign USDG zero collateral factor on Ethereum
Counterparty Low Paxos well-regulated, zero incident history
Overall Medium Acceptable at conservative parameters; note that no major venue currently agrees

Recommended parameters for lending against USDG:

  • Max LTV: 80% — the freeze/wipe capability is the binding consideration: if Paxos freezes a borrower's collateral, the position becomes unliquidatable at any LTV, so the buffer is protecting against an event that LTV cannot actually mitigate. An 80% ceiling keeps the position economically sensible for a same-value pair while acknowledging that every major independent risk manager has declined to lend against USDG at all
  • Liquidation threshold: 85%
  • Max exposure: min($5M, 25% of Ethereum USDG DEX liquidity, 10% of total vault TVL) — the DEX-liquidity fraction is set at 25% because the Curve pool's usable depth is bounded by its ~$18.1M USDC side, not by its ~$30.5M headline TVL
  • Prefer the market-standard shape: if the objective is USDG exposure rather than USDG collateral, supplying USDG as a loan asset (Aave v3/v4, Maple, Morpho) matches how every major venue has chosen to underwrite it
  • Monitor: Paxos freeze events (ASSET_PROTECTION_ROLE), Curve pool USDC-side balance rather than headline TVL, USDG peg, and any change in Aave/Morpho collateral-factor configuration — a venue enabling USDG as collateral would be a meaningful signal to revisit these parameters upward
Score Details

Scoring Guidelines:

  • Be conservative: when uncertain between two scores, choose the higher (riskier) one
  • Use decimals (e.g., 2.5) when a subcategory falls between scores
  • Prioritize onchain evidence over documentation claims

Critical Risk Gates

  • Unverified contract source — token proxy, token implementation, SupplyControl proxy/implementation, all five live facets, and the OFTWrapper are source-verified on Etherscan (confirmed via getsourcecode, August 15, 2026). ✅ PASS
  • No audit — 6 USDG-relevant audits by 3 reputable firms (Zellic, Trail of Bits, Halborn), including one covering the V3 rewards system. ✅ PASS
  • Unverifiable reserves — Offchain reserves, but monthly attestation by KPMG LLP under ISCA standards, MAS regulatory oversight, segregated accounts. Multiple independent verification layers. ✅ PASS (same structure as USDC)
  • Total centralization — Token admin is a TimelockController (24h), and both 2026 upgrades were demonstrably routed through it. Operational roles (PAUSE, ASSET_PROTECTION) held by an MPC wallet (key sharded across multiple parties). SupplyControl admin under timelock. Not fully decentralized but not a single-key EOA. ✅ PASS

All gates pass. Proceed to category scoring.

Category Scores

Expand a category to see how it was scored.

Audits & Historical Track Record20%1.50

Subcategory A: Audits & Security Reviews

Factor Assessment
Audits 6 USDG-relevant audits from 3 reputable firms: Zellic (3), Trail of Bits (1), Halborn (2). Covers core contracts, cross-chain, rewards, and signatures
Bug bounty $1M max payout, live on Cantina since March 27, 2026, Cantina-triaged, USDG contracts and cross-chain infrastructure explicitly in scope. Invite-only rather than open
Contract surface Increased materially — UUPS proxy + 118 facet-routed selectors + 10 roles + rewards share accounting on the transfer path
V3 audit depth 1.2 person-weeks over 4 calendar days for the largest change since launch; private repo, so deployed bytecode is not third-party matchable to the audited commit

Audits Score: 1.5/5 — Rubric row 1 requires "3+ audits by top firms" (satisfied: Zellic ×3, Trail of Bits, Halborn ×2) and "Active, max payout >$1M" (nearly satisfied: exactly $1M, live and triaged, but invite-only). Two factors hold this off a clean 1: the bounty is gated rather than open, and the rubric's note that "simple contract surface scores better than highly complex ones" now cuts against USDG — the diamond-facet rewards system is a substantial new surface reviewed on a light engagement. → 1.5.

Subcategory B: Historical Track Record

Factor Assessment
Production history ~22 months (Oct 7, 2024 deployment). Paxos has 7+ years stablecoin experience (USDP since 2018)
TVL ~$3.47B total supply, ~$449M on Ethereum — sustained well above $100M for over a year
Security incidents None. Zero incidents across all Paxos stablecoins (USDP, PYUSD, USDG)
Peg record Never below $0.995 over the trailing 365 days

Historical Score: 1.5/5 — Scale is unambiguously row 1 (sustained >$100M, by a factor of 30). Time in production is 22 months, just short of the ">2 years" row-1 threshold and at the top of the 1–2 year row-2 band. Splitting these gives 1.5. This will mechanically improve to 1.0 in October 2026 absent any incident.

Audits & Historical Score = (1.5 + 1.5) / 2 = 1.5

Score: 1.5/5 — Six audits from three reputable firms, a live $1M Cantina bug bounty covering USDG, 22 months in production, $3.47B of supply, and zero incidents across any Paxos stablecoin. What keeps this off 1.0 is the combination of an invite-only rather than open bounty and a contract surface that expanded sharply with V3 — 118 facet-routed selectors reviewed on a 1.2 person-week engagement against a private repo.

Centralization & Control Risks30%2.70

Subcategory A: Governance

Factor Assessment
Upgradeability UUPS proxy — upgradeable via TimelockController (24h delay). Both 2026 upgrades verifiably routed through the timelock
Token admin TimelockController (24h) — PROPOSER/EXECUTOR/CANCELLER all held by MPC wallet
Operational roles MPC wallet holds PAUSE, ASSET_PROTECTION directly (no onchain timelock, but MPC provides multi-party internal approvals)
SupplyControl admin DEFAULT_ADMIN under 24h timelock; SCM held by same MPC wallet
Supply controller scope Mitigant: SC1/SC2 restricted to one whitelisted mint destination each and can only burn their own balance — verified onchain, never changed since registration
Bridge configuration Aggravator: MPC wallet is owner() of the OFT wrapper and its LayerZero delegate, with no timelock, upstream of a 200M USDG/24h mint path
Rewards roles Aggravator: 6 new roles since Feb 2026; three held by single-key hot-wallet EOAs; two MPC-held roles can redirect reward payouts
MPC transparency Internal quorum/threshold and policy configuration not publicly verifiable — security depends on Paxos's internal controls
Regulatory MAS-supervised issuer; OCC-supervised US group entity — provides offchain governance and accountability

Governance Score: 3.0/5 — Rubric row 3 reads "some powerful roles, constrained by timelock" with a "24+ hours" delay, and that is an accurate description of USDG. Upgrades, facet changes, role management, and SupplyControl administration all sit behind a 24-hour TimelockController that is demonstrably exercised rather than nominal — both 2026 upgrades were routed through it. Pause and freeze are exercisable immediately, but that is by design: an emergency control that has to wait 24 hours is not an emergency control, and every regulated fiat stablecoin is built this way.

Supporting a 3 rather than a 4:

  • Supply controller scope is genuinely constrained. SC1 and SC2 can each mint to exactly one whitelisted destination and can only burn their own balance, so neither key can mint to an attacker or seize a holder's tokens.
  • The timelock is real. 24 hours of advance notice on every upgrade, with CallScheduled publicly observable, and deep enough exit liquidity for integrators to act on it.
  • MPC rather than a single key, with institutional custody infrastructure confirmed by the gas-station funding pattern.
  • Regulatory accountability from MAS supervision of the issuer and OCC supervision of the US group entity.

Holding it off a 2 (which asks for "limited roles, cannot seize funds"):

  • ASSET_PROTECTION_ROLE can seize funds — freeze and wipe are unilateral from the contract's perspective, which alone rules out row 2 regardless of the regulatory justification.
  • No independent canceller. The MPC wallet is the sole holder of proposer, executor, and canceller, verified against the timelock's complete role history (four grants, zero revocations). A timelock whose only canceller is its only proposer delays a compromised admin; it cannot stop one.
  • Bridge configuration is not behind timelock, upstream of a 200M USDG/24h mint path with unrestricted destinations.
  • Three single-key EOA role-holders on the V3 rewards system.
  • The internal MPC quorum is not publicly verifiable.

Subcategory B: Programmability

Factor Assessment
Token operations Standard ERC-20, fully onchain and programmatic
Exchange rate Fixed 1:1 USD peg — no PPS to compute, no oracle dependency, no admin-updatable rate
Minting/burning Onchain via SupplyControl with onchain rate limits and destination whitelists, but Paxos-controlled (not permissionless)
Partner rewards Now onchain (V3, Feb 2026) — shares/multiplier accrual settled from a funded claim source, replacing 30-day offchain reconciliation
Reserves Entirely offchain — managed by Paxos and custodians
Reporting Monthly attestation (offchain), no real-time onchain reporting

Programmability Score: 3.0/5 — A clean rubric row 3. Transfers, approvals, mint/burn authorization, rate limiting, and reward accrual all execute onchain, with offchain reserve management behind them. On the rubric's second column, PPS/Rate Definition, USDG has no exchange rate for an admin to set — the peg is a hard 1:1 constant, which is better than row 3's "onchain but reliant on admin updates," not worse. Reserve custody is offchain, but that is a collateralization and provability property scored in Category 3, and counting it here would double-count it. Row 4 ("significant manual intervention required," "offchain accounting with periodic reporting") does not describe the token's operations, particularly now that V3 has moved partner reward accounting onchain.

Subcategory C: External Dependencies

Factor Assessment
DeFi protocols None — standalone stablecoin
Infrastructure Banking/custodian (inherent to fiat-backed), LayerZero V2 (bridging)
Criticality Banking dependency is fundamental but not a DeFi protocol risk. LayerZero is non-critical to an Ethereum holder's balance but is now an inbound mint path of up to 200M USDG/24h
Exposure trend ~87% of total supply sits on chains reachable only through the wrapper; SC3 capacity 45M → 200M since the last assessment

Dependencies Score: 2.0/5 — Rubric row 2, "1-2 blue-chip dependencies / non-critical," describes this squarely. There are exactly two — banking custodians and LayerZero V2 — and neither is required for an Ethereum holder to hold, transfer, or exit USDG. The LayerZero exposure is not marginal enough to round toward row 1's "no external dependencies," though: the wrapper is a registered supply controller with unrestricted mint destinations and a 200M USDG/24h ceiling, and its peer set and DVN configuration are controlled by a key with no timelock. The 3-of-3 DVN quorum on every configured route is a real protection, but only two of the three verifiers are independent of Paxos, and the quorum itself is mutable by the wrapper's owner. Row 3 would require "some critical functions depend on them," which is not the case — the bridge can push supply in, but nothing an Ethereum holder needs depends on it.

Centralization Score = (3.0 + 3.0 + 2.0) / 3 = 2.67

Score: 2.7/5 — Governance is concentrated in a single MPC wallet, but the powers that can move or destroy user funds are either behind a 24-hour timelock (upgrades, facets, roles, SupplyControl admin) or scope-limited at the contract level (SC1/SC2 mint whitelisting). Token operations are fully onchain; reserves are not. External dependencies are minimal and non-critical to an Ethereum holder. The unresolved items are the freeze capability, the absence of an independent canceller, and bridge configuration that sits outside the timelock.

Funds Management30%2.50

Subcategory A: Collateralization

Factor Assessment
Backing 100% backed by cash and cash equivalents (U.S. T-Bills, cash)
Collateral quality U.S. Treasury Bills — lowest-risk financial instrument globally
Segregation Reserves held in segregated accounts, bankruptcy-remote
Regulatory MAS requires 1:1 reserve maintenance
Onchain All reserves are offchain

Collateralization Score: 2.0/5 — Highest-quality collateral (U.S. T-Bills), 100% backing required by regulation, segregated accounts. However, reserves are entirely offchain. Per rubric: between score 2 ("100% onchain collateral, high quality DeFi assets") and score 3 ("100% collateral, some offchain"). Quality is the highest possible (U.S. Treasuries) but fully offchain — regulatory framework partially compensates. Same collateral profile as USDC and USTB.

Subcategory B: Provability

Factor Assessment
Reserve transparency Monthly attestation by KPMG LLP (Big Four) under ISCA standards, for all reports since February 27, 2026
Onchain verification Supply verifiable onchain on every chain; reserves are not
Chainlink PoR Not available
Reporting Monthly reports published on the transparency portal
Regulatory oversight MAS supervision of the issuer with regulatory reporting requirements; OCC supervision of the US group entity

Provability Score: 3.0/5 — Rubric row 3 reads "Hybrid onchain/offchain | Manual reporting by admins | Known custodian attestation," which describes the current arrangement precisely on all three columns. Attestations are monthly and performed by KPMG LLP under Institute of Singapore Chartered Accountants standards, so row 4's "infrequent reporting" and "self-reported only" do not apply. Real-time onchain verification is absent and no Chainlink PoR exists, which is what keeps this at 3 rather than 2. Consistent with the corpus: superstate-ustb, whose Treasury holdings sit behind a gated investor portal, also scores 3.0 on provability.

Funds Management Score = (2.0 + 3.0) / 2 = 2.5

Score: 2.5/5 — Highest-quality collateral (U.S. Treasuries) with regulatory requirements for 1:1 backing and segregated custody. The offchain nature of the reserves is the primary weakness, mitigated by monthly KPMG examination under a named professional standard and MAS supervision of the issuer.

Liquidity Risk15%1.50
Factor Assessment
Exit mechanism DEX swap (permissionless), CEX trade, or direct Paxos redemption (KYC, business hours)
DEX liquidity (ETH) ~$34.3M across 8 pools ($30.5M in Curve USDG/USDC)
CEX liquidity OKX, Bullish, Biconomy, Kraken, Gate — ~$11.3M 24h volume; KuCoin does not list USDG
Direct redemption 1:1 from Paxos (unlimited, KYC required)
Same-value asset USD stablecoin — no price divergence risk; never traded below $0.995 in 365 days
Slippage (measured onchain) $10M at 0.023%; $15M at 0.068%; single-pool ceiling ~$17M

Score: 1.5/5 — Scoring the rubric column by column:

  • Liquidity depth — row 1 requires ">$10M, <0.5% slippage." Measured: $10M executes at 0.023% and $15M at 0.068% in a single Curve pool, before any routing across the other seven Ethereum pools or CEX venues. Comfortably row 1.
  • Large holder impact — row 1 requires "full exit with <0.5% impact." Any position up to ~$15M clears that on Curve alone. Row 1.
  • Exit mechanism — direct 1:1 redemption exists and is unlimited, but requires a KYC account and business-hours processing, so it is row 2 ("direct redemption with minor delays") rather than row 1 ("instant").
  • Adjustment — USDG is a same-value asset, which the rubric notes "can accept higher exit times."

Two of three columns are row 1, one is row 2, with a favourable adjustment → 1.5. The residual concerns argue against going to 1.0 rather than against 1.5: depth is concentrated, with 89% of Ethereum DEX liquidity in one Curve pool and a hard cliff past ~$17M once the USDC side is drawn down, and the CEX backstop is thin at ~$11.3M.

Operational Risk5%1.50
Factor Assessment
Team Paxos: founded 2012, 200+ employees. Charles Cascarilla (CEO), Rich Teo (co-founder)
Stablecoin experience 7+ years: USDP (2018), PYUSD (2023), USDG (2024). Zero incidents
Regulation MAS-supervised issuer (Singapore); US group entity OCC-supervised since Dec 12, 2025 (Charter 25379, converted from NYDFS); MiCA compliant
Documentation Comprehensive integration docs and open-source MIT code — but governance and the V3 rewards system are undocumented, and the GitHub README's role list is over a year stale
Legal structure Singapore Major Payments Institution, with a US federal trust bank and EU sister entity
Partners Kraken, Robinhood, Galaxy Digital, BitGo, Anchorage Digital

Score: 1.5/5 — Paxos is one of the most established and well-regulated companies in the stablecoin space, and the legal/compliance column is row 1 by any reading: a MAS-supervised issuer, an OCC-supervised US federal trust bank, and MiCA compliance in the EU. What holds this at 1.5 rather than 1.0 is the documentation column. Row 1 requires "excellent, comprehensive" documentation, and the public record materially misdescribes the system: the GitHub README names multisigs that have held no roles since August 2025, and neither the governance structure nor a rewards subsystem covering ~80% of Ethereum supply is documented anywhere public. Integration docs are strong; governance and protocol-change disclosure is not.

Final Score Calculation

Final Score = (Centralization × 0.30) + (Funds Mgmt × 0.30) + (Audits × 0.20) + (Liquidity × 0.15) + (Operational × 0.05)
            = (2.7 × 0.30) + (2.5 × 0.30) + (1.5 × 0.20) + (1.5 × 0.15) + (1.5 × 0.05)
            = 0.81 + 0.75 + 0.30 + 0.225 + 0.075
            = 2.16
            ≈ 2.2
Category Score Weight Weighted
Audits & Historical 1.5 20% 0.30
Centralization & Control 2.7 30% 0.81
Funds Management 2.5 30% 0.75
Liquidity Risk 1.5 15% 0.225
Operational Risk 1.5 5% 0.075
Final Score 2.2/5.0

Risk Tier

Final Score Risk Tier Recommendation
1.0-1.5 Minimal Risk Approved, high confidence
1.5-2.5 Low Risk Approved with standard monitoring
2.5-3.5 Medium Risk Approved with enhanced monitoring
3.5-4.5 Elevated Risk Limited approval, strict limits
4.5-5.0 High Risk Not recommended

Final Risk Tier: Low Risk (2.2/5.0) — Approved with standard monitoring

USDG benefits from Paxos's established stablecoin track record, highest-quality collateral (U.S. Treasuries), a strong regulatory framework (MAS for the issuer, OCC federal supervision for the US group entity, MiCA in the EU), solid audit coverage backed by a live $1M bug bounty, and deep exit liquidity — a $10M Ethereum exit clears at 0.023% slippage. Onchain control is consolidated into a single MPC wallet with key sharding across multiple parties, and the 24h timelock on upgrades is demonstrably exercised rather than nominal.

Where the residual risk sits. Liquidity is not the binding constraint, and reserve quality and issuer standing are strong and independently attested. What remains is concentrated in the powers that sit outside the timelock, and it is worth being precise about which of those are informative:

  • Freeze is the largest risk by expected loss for any single integrator — a frozen vault or strategy loses access to its entire USDG position with no onchain recourse. It is mandated by USDG's regulatory posture and identical in kind to USDC's, so it is a reason to size positions, not a reason to prefer another regulated fiat stablecoin. Pause is in the same category and is deliberately immediate, as an emergency control must be.
  • Bridge configuration outside the timelock is the risk specific to USDG. One MPC-controlled address can add a peer chain and rewrite the DVN quorum with immediate effect, upstream of a supply controller that can mint 200M USDG per day to arbitrary addresses. No regulator requires this, no peer of comparable size is configured this way, and — because the same address is the timelock's only canceller — there is no independent party who could interrupt it. It is the only path in the system that produces unbacked canonical supply, which harms every holder and the peg rather than one address.

Deep exit liquidity is what makes the 24h timelock genuinely useful, since integrators can act on the warning it provides. It also marginally raises how much an attacker could extract per unit time in a mint scenario — though at ~$17M of usable Curve depth against a 200M/24h mint ceiling, exit liquidity remains the binding constraint by roughly an order of magnitude.

For an integrator, the practical implications are to monitor PeerSet, OwnershipTransferred, and DVN-config changes on the wrapper at least as closely as timelock events; to treat any first-ever flip of allowAnyMintAndBurnAddress on SC1 or SC2 as a high-priority alert; and to treat the grant of CANCELLER_ROLE to an independent party as the single highest-value governance improvement Paxos could make.

Assessment History

DateScoreNotes
March 20, 20262.4Initial assessment
June 26, 20262.4Reassessment: governance restructured to a single MPC wallet, timelock 3h → 24h; score unchanged
July 30, 20262.4Reassessment: supply/chain refresh (Robinhood Chain added, 6 chains with material supply), LayerZero peer set and DVN quorum enumerated onchain; score unchanged
August 15, 20262.2Full review, resolving #388. Corrections: US regulator NYDFS → OCC (Paxos Trust Company, N.A., Charter 25379, Dec 12 2025); the supply-controller Allow Any Address column was inverted (SC1/SC2 are whitelist-bound and can only burn their own balance); SC3 bridge capacity is 200M USDG/24h, not 45M; the LayerZero Labs DVN address was malformed. New findings: the previously-unreported V3 rewards system (5 facets, 118 selectors, 6 roles, 28 payout groups holding ~80% of Ethereum supply, no new mint authority); MPC ownership of the OFT wrapper and its DVN config, not behind timelock — the largest residual risk; and no independent canceller on the timelock (MPC wallet is sole proposer/executor/canceller), making the 24h delay a monitoring and exit window rather than a circuit breaker. Refreshed: supply to $3.468B read onchain, DEX/CEX liquidity, peg history, and slippage replaced with onchain get_dy measurements ($10M at 0.023%). Appendix B: every major Ethereum venue assigns USDG a collateral factor of zero, so max LTV cut 90% → 80%. Audits 2.0→1.5, Centralization 2.8→2.7 (Governance 3.5→3.0, Programmability 3.5→3.0, Dependencies 1.5→2.0), Funds Mgmt 2.75→2.5, Liquidity 2.0→1.5; Operational held at 1.5
Submit report inaccuracy