AUSD — Risk Report

Moderate risk · 6.0/10

A small dollar with big-league backing and unusually good transparency. AUSD is Agora’s fiat-backed stablecoin — reserves managed by VanEck and custodied at State Street, held in a bankruptcy-remote structure, backed only by cash, overnight repos and short-term US Treasuries, with monthly PwC attestations and real-time on-chain proof-of-reserves from Chaos Labs. On backing quality and transparency it punches well above its weight. What holds it to a 6.0 rather than higher: it’s small (~$180M), young (~2 years), regulated in Bermuda rather than under a US regime like NYDFS, and its reserves run through a single manager and a single custodian — a concentration real enough that one institutional custodian, Anchorage, delisted it. There’s also a structural nuance worth knowing: the plain AUSD you hold has no cross-chain bridge risk, but its cross-chain wrapper (AUSD0) does route through LayerZero.

Yield	Exit method	Primary redemption	Age	Chains
None	Zero-fee instant swap AUSD ↔ USDC/USDT on Agora; sell on a supporting CEX/DEX; institutional 1:1 fiat redemption	Instant atomic swap to USDC/USDT (everyone) + direct fiat redemption (whitelisted institutions)	Since mid-2024	Native-minted on 17+ chains (Ethereum, Solana, Arbitrum, Avalanche, Base, Polygon, BNB, Sui, Monad…)

Summary

AUSD is a fully fiat-backed stablecoin issued by Agora Bermuda Limited, backed 1:1 by cash, overnight reverse repos and short-term US Treasuries. Reserves are managed by VanEck and custodied by State Street in a bankruptcy-remote, segregated structure, with monthly PwC attestations and on-chain proof-of-reserves via Chaos Labs. It launched in mid-2024 and sits at roughly $181M across 17+ chains.

The 6.0/10 reflects genuinely strong, transparent backing offset by small scale, a roughly two-year record, Bermuda (non-US) regulation, single-manager / single-custodian concentration, and an upgrade path with no notice period on any measured chain. It’s worth stating the anchor plainly: AUSD sits a full notch below USDT’s 7.0, and it gets there from the opposite direction. USDT earns its number on deep, everywhere liquidity despite weak issuer transparency; AUSD’s case is strong backing transparency against thin liquidity. The shapes are still mirror images and the gap is now a notch rather than a rounding — the contrast is measured, not rhetorical: AUSD backing 8.0 against USDT’s 6.0, USDT liquidity 9.5 against AUSD’s 6.5.

1 · Stability — 8.5

Here’s the reason strong backing doesn’t carry the score higher. AUSD’s reserves depend on a single manager (VanEck) and a single custodian (State Street). There is no diversification: if either firm had an operational failure, a legal freeze, or lost access, all of AUSD’s backing would be affected at once. There’s no second manager or second custodian to fall back on.

This isn’t a hypothetical concern that only analysts raise. Anchorage Digital delisted AUSD in June 2025, citing exactly this concentration. Anchorage is a conservative, regulated custodian, so that’s an independent counterparty signal worth weighing — not proof that anything is wrong with the reserves, but a real institution deciding the concentration was more than it wanted to hold.

2 · Backing — 8.0

AUSD is backed 1:1 by cash, overnight reverse repos, and short-term US Treasuries — and nothing else. No unsecured commercial paper, no money-market fund shares, no long-duration bonds. This is about as conservative as a fiat-reserve mix gets, and it means the reserves don’t carry meaningful interest-rate or credit risk of their own.

The reserves are managed by VanEck (a >$100B asset manager) and custodied by State Street ($4T+ under custody), in a bankruptcy-remote, segregated account. That structure is designed so AUSD holders’ claim on the reserves survives even if Agora itself failed.

What sets AUSD apart from most stablecoins its size is two layers of transparency, not one: monthly PwC attestations and real-time on-chain proof-of-reserves via Chaos Labs. The on-chain PoR is the standout — it’s a live, verifiable feed rather than a point-in-time PDF. That’s a step beyond USDT (which publishes no on-chain PoR) and even beyond USDC (attestations, but not real-time on-chain proof). This is why AUSD’s backing axis scores 8.0 — the reserves are clean and the proof is genuinely good.

3 · Liquidity & Exit — 6.5

AUSD’s headline exit is a zero-fee, atomic instant swap to USDC or USDT on Agora’s Stable Swap protocol. This is a very strong peg mechanism: anyone can swap AUSD for USDC/USDT at 1:1 with no fee, instantly, on-chain. That’s the tightest peg-arbitrage leash of any sub-$1B stablecoin, and it’s what backfills the token’s otherwise modest DEX depth — if AUSD ever traded below a dollar, arbitrageurs can swap it back to par all day.

On top of that, whitelisted institutions can redeem 1:1 directly to fiat.

The caveat: secondary market depth on DEXes is thin on most chains (moderate on Ethereum). So in a stress scenario where the instant-swap mechanism were paused, secondary liquidity would not absorb large forced sellers cleanly. The practical rule is to size your position against the assumption that the swap mechanism is live — because when it’s live, exit is excellent, and when it isn’t, the open market is shallow.

4 · Dependencies — 7.0

✅ The service-provider set is genuinely diversified, and this is the strongest axis on the report. Reserves are managed by VanEck, custodied at State Street, attested by PwC, and risk-monitored by Chaos Labs — four distinct firms with no role overlap and no circular legs, where one firm marking its own work is the usual failure.

⚠️ What holds it below higher is that there is no redundancy at either layer. One manager, one custodian. The set is diverse across functions but single-threaded within each, so the axis prices breadth of counterparty type rather than depth of substitution.

⚠️ This overlaps with a dock already taken on Backing, where the 8.0 is reduced for single-manager and single-custodian concentration — the same facts. The overlap is flagged rather than resolved here, because re-deriving Backing is a separate judgement from publishing this one.

5 · Contract & Admin — 2.5

⚠️ Upgrade authority on AUSD is a single address with no on-chain quorum and no delay, and it is the same address on every EVM chain it is deployed to. Measured 2026-09-10 across all eight EVM deployments — Ethereum, Arbitrum, Avalanche, Base, Polygon, BNB Chain, Mantle and Monad:

read	result
token `0x00000000eFE302BEAA2b3e6e1b18d08D69a9012a`	identical 5,937-byte proxy on all eight
EIP-1967 admin slot	`0xb8fcc66d613e5f54ee6a425ddbf4a2fdbe4dedee` — the same ProxyAdmin on all eight
`ProxyAdmin.owner()`	`0x68898B77EbF7b55dCA8A2e62d6Fd74959a2930e2` on all eight
`getMinDelay()` on that admin	reverts on all eight — there is no TimelockController
`eth_getCode` on the owner	`0x` on all eight — no contract, so no on-chain quorum

Two fabricated control addresses revert on the same calls, so the reads discriminate.

⚠️ What that means, stated precisely. upgradeAndCall is gated on onlyOwner, the owner is an account with no code, and no timelock sits in front of it. So a single signature replaces the token implementation, and it takes effect immediately. ✅ Ownable2Step is present, and it does not help here — it makes transferring ownership two-step; it does nothing to constrain using it.

⚠️ And the same address controls all eight chains at once, because an EOA address derives from a public key and is chain-independent. One signature is not one chain’s worth of exposure.

✅ What cannot be read on-chain, and is therefore not claimed: whether that address is backed by a single private key or by an off-chain MPC or HSM quorum. eth_getCode returning 0x proves there is no on-chain quorum; it says nothing about custody arrangements that live off-chain. The finding is the absence of an enforceable delay and of any quorum a third party can verify — not a claim about how Agora holds the key.

The comparison that sets the score is USDC, which has the same proxy shape and is credited for using three different owner keys across three chains, so a single compromise reaches one chain. AUSD is the inverse of that design.

⚠️ Coverage: eight of the deployments were read. The two non-EVM legs — Solana and Sui — cannot carry this proxy shape at all and were not assessed here. A ninth address on Blast returned no code, so AUSD does not appear to be deployed there under the same address. Chains not read are unread, not clean.

6 · Issuer — 6.0

Agora is doxxed and well-capitalized: a Paradigm-led $50M Series A, VanEck and State Street as strategic partners, and a team that includes ex-MakerDAO engineers. That’s a credible foundation.

The offsets are real, though. Agora is only about two years old, and it’s regulated under Bermuda’s Segregated Accounts Companies (SAC) Act rather than a US regime like NYDFS. The Bermuda structure is genuinely bankruptcy-remote, but it’s a less-tested enforcement pathway than the US trust structures behind USDC or PYUSD. Standard centralized controls apply — Agora can mint, burn, and freeze the token. For these reasons AUSD scores below NYDFS-regulated peers like Paxos (PYUSD) and Ripple (RLUSD) on the issuer axis.

What you actually earn

Nothing native — plain AUSD pays no yield. You can earn on it through DeFi (for example, AUSD Pendle pools on Monad), but that’s a separate venue decision with its own risks, not a property of the token itself. Holding AUSD is for people who want a clean, transparent dollar, not a yield instrument.

The two-token structure — AUSD vs AUSD0

This is the one structural nuance every holder should understand, and it’s simpler than it sounds. There are two forms of the token:

Plain AUSD — the one you normally hold. It is native-minted on each chain, meaning it’s issued directly on Ethereum, on Solana, on Arbitrum, and so on. It has zero cross-chain bridge exposure: it doesn’t connect to any bridge, so no bridge exploit can forge it. This is the cleaner, safer thing to hold.
AUSD0 — a separate LayerZero cross-chain wrapper used for programmatic bridging between chains. AUSD0’s bridge security has been measured and it passes the current bar for what a bridge should require (multiple independent verifiers, using its strongest configuration on its highest-supply routes). It’s acceptable — but it is still a bridge, and bridges are a distinct risk class.

Retail takeaway: hold plain AUSD, not AUSD0, unless you specifically need to move value cross-chain programmatically. Most holders never need AUSD0 at all.

A note on chain concentration

A large and growing share of AUSD now lives on Monad — roughly 40% of supply as at this report’s 2026-07-08 pass, where AUSD is the largest stablecoin, boosted by Pendle yield pools. Treat that share as indicative rather than current: it is a ratio of two figures that both move, Monad is among the AUSD deployments this coverage cannot read directly, and a chain share can drift materially without either underlying number looking unusual. Monad is a young chain. The relevant risk there isn’t a bridge exploit (that route rides AUSD0’s strongest-secured configuration) but chain liveness: if Monad halted, bridging that supply back to Ethereum could be delayed. It’s a “how young is the chain” consideration, not a “can the bridge be hacked” one.

Track record

Roughly 26 months, and clean: no depegs, no exploits on AUSD, AUSD0, or the mint/redeem path, and no regulatory actions. That’s a reassuring record.

The honest caveat is that this record was built in a calm regime. AUSD has not yet been tested through a major stress event — a USDC-style banking crisis, or a problem specific to VanEck or State Street. A clean two-year run in good conditions is a positive signal, but it’s not the same as multi-cycle, tested-under-fire proof.

Audits & security

AUSD’s canonical token has been audited by top-tier firms — Cantina/Spearbit, with Certora formal verification, plus Zellic/MoveBit for the Sui build — and carries an Immunefi bug bounty. The AUSD0 wrapper inherits LayerZero’s audited reference implementation; Agora’s AUSD0-specific audit scope is less clearly documented. The residual risks here are issuer, concentration and scale — not contract bugs in the canonical token. But “not a contract bug” is not the same as “not a contract risk,” and the upgrade path is worth reading separately.

The upgrade path has no timelock, on every chain measured. AUSD is an OpenZeppelin v5 transparent proxy, and on all nine chains checked — Ethereum, Arbitrum, Avalanche, Base, BNB, Fraxtal, Gnosis, Monad and Polygon — the ProxyAdmin is owned by a single externally-owned key with no delay anywhere in the path. Signing threshold on-chain is one, and there is no queued-action window: an upgrade can replace the token’s logic immediately, with no notice a holder could act on.

⚠️ Correction, 2026-08-24 — this report published the opposite of the truth on the containment question, and the correction is not in Agora’s favour.

Until today this section said Agora used “a different key per chain, which is genuine containment: a single compromise reaches one deployment rather than all nine.” That is false. Read across every chain reachable in one batch on 2026-08-24, six of six return the same ProxyAdmin (0xb8fcc66d613e5f54ee6a425ddbf4a2fdbe4dedee) and the same owner (0x68898B77…30e2), with zero bytes of code at the owner: Ethereum, Arbitrum, Base, Polygon, Fraxtal and Avalanche.

⚠️ Scope that precisely: six chains measured on one key, not nine. Gnosis, BNB and Monad were not reached in that batch, so their owner is not confirmed to be the same key — the separate finding that every measured chain lacks a timelock rests on the earlier nine-chain read and is unaffected.

⚠️ And for an externally-owned account, the same address across chains is not a coincidence — it is necessarily the same private key. An EOA’s address is derived from its public key, so it cannot be independently deployed the way a contract can. (This is the reverse of the situation with contracts: the same contract address on two chains is two separate deployments with potentially different owners, which is why this coverage checks them per-chain. For a plain key, one address means one key, full stop.)

A single compromise reaches at least six chains, not one — so there is no containment benefit to credit here.

⚠️ One trap worth naming, because it is what nearly kept a softened version of this alive: the Avalanche ProxyAdmin’s bytecode genuinely does differ from its siblings. That is true and it is irrelevant. A distinct contract can be owned by an identical key — distinct deployment, single key. Codehash and ownership are separate variables, and reading a differing codehash as evidence of a differing key is the same conflation in miniature as reading a shared address as evidence of a shared owner.

What the Ethereum leg alone looks like, measured: the token holds $68.19M of supply behind an OpenZeppelin v5 ProxyAdmin whose owner() is that bare key, with pendingOwner() at the zero address. ⚠️ Ownable2Step is present and does not help — it makes transferring ownership a two-step process; it does nothing to constrain using it.

⚠️ State the quantifier before the count: this is any-of-one, not one-of-n. There is no quorum here that a single key happens to be a member of. There is one unilateral power, exercisable in a single signature, with no queued-action window.

The comparison that makes this legible is USDC, which has the identical shape and the opposite topology. Circle’s upgrade authority is also a bare externally-owned key in a legacy admin slot with no delay — this coverage docked it half a notch for exactly that. But Circle uses three different keys across three chains, so one compromise reaches one chain. Agora is the inverse: one key, five chains. Same primitive, opposite blast radius.

One limit belongs with all of it, and it cuts the other way. A key with no code is a statement about the chain, not about Agora. Off-chain controls around that key — an HSM, an MPC scheme, a policy engine requiring multiple approvals before it signs — are invisible from here and would change the picture materially. The key is single on-chain. That should not be softened on the strength of Agora’s reputation, and it should not be hardened past what a chain read supports.

⚠️ Why the score does not move again on this. The Issuer axis was cut to 6.0 earlier on 2026-08-23 for “a bare externally-owned key owning the ProxyAdmin on all nine chains measured, with no timelock anywhere in it” — the axis already priced nine-chain reach. It was the body that contradicted the axis, by crediting a containment the axis had never assumed. Correcting the prose brings the two into line rather than establishing a new drag. What would move it further is the reverse: Agora demonstrating that the key sits behind a real approval scheme, or routing the path through a timelock.

This finding is now priced. When first published on 2026-08-23 it was documented but deliberately left out of the score, pending a pass that could rank it against comparable findings elsewhere in this coverage rather than scoring it on the one report we happened to measure first. That pass has since run, and the Issuer axis moved 6.5 → 6.0 on it. The discriminating question turned out not to be how many signatures an upgrade needs, but whether a holder gets a reaction window: USDS has the same upgrade capability and is not docked, because its path carries a verified 48-hour delay with a cancellation window — though ⚠️ as of 2026-08-25 that window is known not to be independent, since schedule and cancel there share one authority, making it notice a holder can act on rather than an interruption a third party can apply; thBILL has a 3-of-5 quorum and is docked, because it has no notice period at all. AUSD has no delay on any of the nine chains measured, so it falls on the docked side regardless of how the key is held.

Score breakdown

Dimension	Score	Notes
Peg Mechanism	8.5	Direct 1:1 institutional fiat redemption plus a zero-fee, atomic AUSD↔USDC/USDT instant swap open to everyone — a very strong arbitrage leash. AUSD’s peg does not depend on the AUSD0 bridge.
Backing	8.0	Cash + reverse repo + short-term Treasuries only; VanEck-managed, State Street-custodied, bankruptcy-remote; monthly PwC attestations plus real-time on-chain PoR (Chaos Labs). Top-tier transparency for its size; docked for single-manager / single-custodian concentration and Bermuda jurisdiction.
Liquidity	6.5	~$181M cap; CEX presence (Kraken/Bybit/MEXC) but not Coinbase/Binance; DEX depth moderate on Ethereum, thin elsewhere. The zero-fee instant swap backfills materially, but observed secondary depth is limited and would strain under a stressed large exit.
Issuer	6.0	Doxxed, well-capitalized (Paradigm Series A; VanEck/State Street partners), ~26-month clean record. Below the NYDFS-regulated cohort (Paxos/Ripple): Bermuda-SAC regulation, short track record, single-manager / single-custodian concentration (Anchorage delisted over it), and AUSD0 documentation opacity. Priced on the upgrade path documented above: a bare externally-owned key owning the ProxyAdmin on all nine chains measured, with no timelock anywhere in it. See the cohort note below the table.
Overall	6.0	Follows the Issuer axis. Genuinely strong, transparent backing (on-chain PoR, cash + T-bills, VanEck/State Street) for a newcomer, with a clean record and a bridge-exploit class structurally avoided on canonical AUSD — offset by small scale, ~2-year history, Bermuda (non-US) regulation, reserve-manager concentration, thin secondary liquidity, and now an undelayed upgrade path. Overall follows Issuer here because on this report Overall tracks the weakest axis — it sat at 6.5 when Issuer and Liquidity were both 6.5, and Issuer is now the floor at 6.0. (Contrast USDT, where Overall is liquidity-weighted and sits above several of its own axes, so an Issuer move there does not carry the composite.) Sits a notch below USDT’s 7.0, from the opposite direction: strong backing transparency / thin liquidity, versus Tether’s deep liquidity / weak issuer transparency.

Who it’s for

Holders who value backing transparency — on-chain PoR, clean cash + T-bills reserves, VanEck/State Street — and the zero-fee instant swap, who are comfortable with a smaller, younger, Bermuda-regulated issuer, and who hold plain AUSD on a major chain rather than the AUSD0 bridge wrapper.

Who should avoid

Anyone needing deep, everywhere liquidity — that’s USDC or USDT, not AUSD.
Anyone who requires US (NYDFS-style) regulation, or diversified reserve managers and custodians — the single-manager / single-custodian setup is a real concentration.
Anyone needing a censorship-resistant dollar — AUSD is a freezable, centralized-issuer token.

What to watch

Reserve-manager / custodian diversification. The single-VanEck / single-State-Street concentration is the core structural risk. Diversification, or a re-list by Anchorage, would be a positive signal; further conservative delistings, a negative one.
The AUSD↔AUSD0 swap and AUSD0 bridge config. Hold plain AUSD; if you do use AUSD0, its cross-chain security configuration is the thing to track.
Monad concentration. A large, growing share of supply sits on a young chain.
The first real stress test. AUSD hasn’t been through a banking-crisis-style event; how the instant swap and redemption hold up under one is the open question.
Regulatory progress. Agora pursuing US money-transmitter licenses would strengthen the issuer picture.
A timelock appearing on the upgrade path. Today the ProxyAdmin is owned by a plain key with no delay on all nine chains measured. Agora routing that through a timelock or a verifiable multisig would be visible on-chain, and it is the change that would most improve the admin picture described above.

This report is based on Agora’s public documentation, PwC attestations, Chaos Labs on-chain proof-of-reserves, and on-chain data through 2026-07-08, with the upgrade-path read taken 2026-08-23. AUSD is a smaller, younger, Bermuda-regulated, centralized and freezable issuer token with single-manager / single-custodian reserve concentration; its supply, chain distribution, and the AUSD0 bridge configuration all shift over time. Corrections or attestation links welcome at info@tidresearch.com.

Revision history

2026-08-24 — chain reach measured. ⚠️ Six of six chains measured share the owner 0x68898B77…30e2 via ProxyAdmin 0xb8fcc66d613e5f54ee6a425ddbf4a2fdbe4dedee, which holds no code. A single compromise reaches at least six deployments, and Gnosis, BNB and Monad were not reached in that batch. The upgrade path has no timelock on every chain measured.
2026-08-23 — Issuer 7.0 → 6.5, Overall 7.0 → 6.5 on admin-cohort recalibration.
2026-07-08 — Issuer and Overall 7.5 → 7.0 on cohort anchoring. Initial production publication.