Pharos risk

TL;DR

Pharos is an EVM Layer-1 aimed at real-world assets, built by leadership and engineers from Ant Group. Its public mainnet and the PROS token went live on 2026-04-28. It is a chain, not a stablecoin. Pharos issues no dollar of its own: the USDC on it is Circle's own deployment, moved in and out through CCTP. This report rates the chain as a place to hold money. It does not rate the credit vaults Pharos distributes, and those vaults are where the advertised 13 to 15% comes from.

The security work is better published than on most networks this young. Reports from Zellic, ExVul and OpenZeppelin sit in Pharos's own GitHub and cover consensus, execution, cryptography and the system contracts. There are 31 live validators, most of them run by known staking firms, and no outside operator runs more than 6.05% of stake.

Three things keep it at caution. First, the admin keys. A 3-of-4 Safe holds the top admin role on the system contracts and carried out both upgrades, but a plain account that is one of its signers still holds the same role on the Staking and ChainConfig contracts by itself. That role can upgrade them, pause staking and rewrite chain parameters, and we found no timelock in front of it. Second, the stake. Five Safes whose signers overlap with the foundation's supplied 99.8% of the PROS behind those 31 validators, there is no slashing, and the audited design gives the admin role a protocol-level address blacklist. Third, 86% of PROS has not reached the market yet.

Checklist

Audits & contracts. Pharos keeps its audit reports in a public GitHub repo, PharosNetwork/pharos-audit, added on 2026-04-14. Zellic wrote four, totalling 31.4 person-weeks:

• cryptography and precompiles (2025-08-18): 1 critical, 2 high, 3 medium, 10 low, all marked fixed
• consensus and P2P networking (2025-08-28): 15 critical, 2 high, 2 low. Fourteen criticals fixed, one partly fixed, one high acknowledged
• the execution layer (2025-09-04): 7 critical, 2 high, 3 medium, 1 low. Three criticals fixed. Pharos disputed the other four, and one high, with written rebuttals
• a PoS update (2025-09-30): 2 high, 1 low. One high, on test coverage, acknowledged only

In the first three reports Zellic wrote that "based on the number of severe findings uncovered during the audit, we strongly advise a comprehensive reassessment before deployment". ExVul reviewed the node and system code from July to December 2025, finishing after Zellic. It found 17 critical, 15 high, 21 medium, 34 low and 14 informational issues. Every critical, high and medium is marked fixed, and seven lows and one informational are acknowledged. OpenZeppelin re-audited the system contracts in January 2026 (report dated 2026-02-09) and found 27 issues: 4 high, none critical, all marked resolved.

That covers the parts of a chain that lose money when they break. The qualifiers are real, though. Zellic and ExVul between them logged 40 critical findings before launch. Thirty-five are marked fixed and one partly fixed, and Pharos disputed four, all in the execution layer, on parallel-execution conflict tracking and gas accounting. That describes a young codebase. The node repository the audits cite, AntChainAldabaNG, is not public (GitHub returns 404). Pharos did publish its consensus library (PharosTumbler) and networking library (PharosCubenet) under Apache-2.0 in April 2026, but nobody outside can match the audited commits to what validators run. Staking and ChainConfig were also upgraded after genesis. The new Staking code carries the delegation system and V1-to-V2 migration that OpenZeppelin's re-audit covered, and both new implementations drop functions (addStake, updateImplAddress) that OpenZeppelin's findings flagged and the published source no longer has. That fits the re-audited code, but we could not match bytecode to the audited commit. R25, the largest app on the chain, shows no audit on DeFiLlama. Press coverage credits SlowMist with auditing its vault contracts; we found no public report.

Admin control. We read this on-chain today through rpc.pharos.xyz. At least three accounts hold DEFAULT_ADMIN_ROLE, the top role, on Staking (0x4100...0000) and ChainConfig (0x3100...0000):

• a Safe v1.4.1 at 0xcce1...31d0 that needs 3 of 4 signatures, granted the role on 2026-03-02
• the protocol's system sender, 0x1111...1111, which advances epochs
• a plain account, 0x5f09...d218, with no contract code, a nonce of 68 and about 82,051 PROS. The genesis file funded it with 100,000 PROS, it has held the role since the first block, and it is one of the Safe's four signers

On 2026-03-02 the Safe revoked this account's role on RuleManager (0x2100...0000), whose admins are now the Safe and the system sender. The account still holds it on Staking and ChainConfig. Any single holder can use the role, so one key can still do alone what the Safe needs three signatures for.

OpenZeppelin's report lists what the role controls across the system contracts:

• upgrades
• minimum stake, epoch length, withdrawal windows and inflation
• pausing staking
• pointing Staking at a different config contract
• the rules that govern transaction validation (RuleManager, now held by the Safe)

The ChainConfig switch that turns PoS rewards on, the route by which new PROS is minted, sits under the same role. The report calls the admin "an off-chain multisig wallet". The plain account has no code, so if it is a threshold wallet, its signers cannot be checked from the chain. Nothing we found sits between a decision and an upgrade, and the power has been used. On 2026-04-14, two weeks before the public launch, the Safe upgraded Staking to 0x874e...04d5 (with a V1-to-V2 migration call) and ChainConfig to 0xd3be...9148. Neither is in the genesis file. RuleManager still runs its genesis code.

The validator set is wider than the money behind it:

• 31 active validators, none pending, with 115,609,950 PROS staked.
• The four genesis validators, now labelled PharosFoundation0 to 3, hold 5,000,000 PROS each (17.30% together). They share one owner, a Safe v1.4.1 at 0xfa0a...6549 that needs 3 of 5 signatures.
• Four more, served from zan.top, belong to ZAN, one of the RPC providers listed in Pharos's docs, with 3.46%. The other 23 are staking firms such as InfStones, B-Harvest, Stakely, Nodes.Guru and HashKey Cloud.
• The operators put up almost none of it. 26 of the 27 non-foundation validators carry 100 PROS from their own owner, and the 27th, Sunthycloud, 207,300. The other 95.4M PROS was delegated by four Safes that need 2 of 3 signatures each, and on every one of them two of the three signers also sign for the foundation's Safe.

So five Safes with overlapping signers supplied 99.8% of all staked PROS. Pharos's published consensus library says voting weight is proportional to stake and quorum checks are stake-weighted. By stake, the top seven validators pass one third and the top sixteen pass two thirds, but whoever controls those Safes can withdraw or move the stake, so the real weight sits with them. The docs say there is "currently no automated slashing mechanism", and in the published Staking source the slashValidator function is an empty placeholder. Misbehaviour goes to "governance procedures" that the docs do not describe.

Freeze / seizure. The audited system contracts include TransactionDeny, a protocol-level blacklist the admin can add any address to. OpenZeppelin notes that balances cannot be confiscated, but every future transaction from a listed address is rejected, which is a freeze in practice. We did not find a deployed TransactionDeny on mainnet this run. It is not in the genesis file or at the system addresses we probed (refresh). USDC on Pharos is Circle's contract, so Circle's usual freeze powers apply on top.

Oracle. A chain has no oracle of its own. The question is what the apps on it price against. Pharos's docs list Chainlink Data Streams (with a verifier proxy live on mainnet), Chainlink push feeds for PROS, BTC, ETH, the main stablecoins and several yield-token rates, and Supra's pull oracle. That is a normal menu, and Morpho's WPROS-collateral market on Pharos uses Chainlink's PROS/USD feed. The exceptions are the credit and custody tokens. APC3M, the Axil Prime Credit vault token, is accepted as Morpho collateral at 77% and 86% LLTV and priced from the value its vault reports, about 1.025 USDC per token today. BTCvc, a token its issuer says is backed by custodied BTC, is priced at BTC parity by a custom oracle contract. More broadly, R25's vault tokens are priced from the value the manager reports for off-chain loans, and no oracle checks that number.

Liquidity & exit. For plain USDC the exit is good. Circle's own address list includes Pharos, and the token at 0xC879...1815 is Circle's deployment. CCTP V2 (domain 31) moves it by burn and mint rather than through a bridge pool, and Circle marks Fast Transfer as unnecessary on Pharos because standard attestation is already fast. Chainlink CCIP is also live, with lanes to Ethereum, Base, Polygon and Jovay, and so is LayerZero (endpoint ID 30407).

The depth is thin. About 9.13M USDC exists on the chain today, DeFiLlama counts $12.33M of stablecoins in all, and the largest DEX it tracks there, FaroSwap, holds $1.22M. The Axil Prime Credit vault alone reports $46.4M of assets, about five times the USDC on the whole chain, so most of what its depositors paid in sits in the off-chain loan book. The yield products also run on their own clock. Axil Prime Credit locks deposits for three months, and redemptions can take up to 20 more days. If no redemption request goes in during the window, the deposit rolls into a new three-month cycle. A chain halt would stop both routes until validators restart it.

Yield: real vs emission. Staking pays nothing yet. The live ChainConfig still reads chain.enable_pos_staking = false, which the Staking code uses to turn PoS rewards off. The published plan is zero inflation for six months after launch, then 5% a year from month seven, so new PROS would start reaching stakers around late October 2026. The live Staking contract does not match the second half: its inflation rate reads 9125, which the published source defines as 9.125% a year. Which rate applies when rewards switch on is refresh.

The yields that draw attention come from credit vaults. Axil Prime Credit targets about 14.3% gross. R25 says that figure includes launch incentives (a $300,000 campaign paid in PROS to subscribers in the 15 to 19 July window), and Binance Wallet lists the vault at 13%. The earlier pAlpha vault was reported as 70% Axil consumer credit and 30% Centrifuge Treasury bonds, and its team said the yield blends asset income with direct treasury support. The credit part is real interest from small consumer loans in emerging markets, with Thailand, the Philippines, Indonesia, Pakistan and Mexico named. Losses are reported by the manager, not visible on-chain. DeFiLlama's yield tracker carries only two Pharos pools, both Centrifuge funds. The larger, a Janus Henderson Treasury fund, pays 4.00% on about $4.6M.

Holder concentration. CoinGecko shows 135.6M PROS circulating of 1B (13.56%), about a $95M market cap against about a $700M fully diluted value. The published split:

• team 20% and investors 20%, locked for 12 months from the April 2026 TGE, then vesting over 36 months
• foundation treasury 16% and Lab Co. treasury 9%, released over 48 to 60 months
• ecosystem and community 21%, including a 6% airdrop (1% unlocked at TGE)
• node and liquidity incentives 14%

Insider unlocks begin around April 2027. On-chain, the genesis file put 995,900,000 PROS, 99.59% of supply, in a single plain account (0x00A9...c405). That account holds 24,550,301 today, and we did not trace where the rest went this run. Staked PROS (115.6M) equals about 85% of CoinGecko's circulating figure, and 99.8% of it came from the five Safes described above, so it looks like treasury allocations put to work rather than bought tokens. Whether CoinGecko counts it as circulating is refresh. The token hit its high of $1.13 on launch day, fell to $0.318 on 2026-07-21 and trades near $0.70, up about 71% in 30 days.

Track record. Short and clean. The chain's own clock (epoch 1,760, four hours each) dates genesis to 2025-12-12, and the public mainnet and TGE followed on 2026-04-28. DeFiLlama's hacks list has nothing for Pharos or for the apps we checked on it (R25, FaroSwap, Zona, OpenFi, Brokex, AquaFlux, Bitverse). Web and X searches through today found no halt, rollback, exploit or depeg. The money is newer than the chain: DeFiLlama has Pharos at $0.71M on 2026-07-03, $18.96M on 2026-09-01 and $74.98M today, its highest reading so far. About $48M of today's figure is a single position: 565 BTCvc posted as Morpho collateral, up from 138 in early September. The token's Pharos contract is owned by a plain account, the same one that owns the Vishwa Morpho vaults. Most of the TVL has been there less than a month.

Worst case

There are two, and they need different defences.

The first is the admin keys. Whoever controls 0x5f09...d218 can upgrade Staking, which holds 115,610,133 PROS (about $81M), or rewrite chain parameters without the Safe's other signers, and nothing we found would make them wait. Signer keys shared with the foundation's Safe can also move nearly all the stake that sets consensus weight. The audited design adds a blacklist that can stop any address from transacting. For USDC on Pharos the realistic damage is a freeze or a halt rather than theft, because Circle's contract is not Pharos's to upgrade. Still, being unable to leave during a bad week is a loss of its own.

The second is the credit. Once you take the advertised yield, your Circle dollar becomes a claim on emerging-market consumer loans. The operator values it. It sits in custody the operators describe (Safe and Fordefi with split signers, for pAlpha) but that we could not verify. And it is locked for three months plus up to 20 days. Rising defaults would show up as a NAV cut you cannot exit ahead of. We found no insurance fund or loss reserve, so position size is the mitigation.

Bottom line

Caution. As a chain, Pharos has done more visible security work than most launches its age: three audit firms with published reports, 31 validators mostly run by known operators, and a native Circle USDC exit. What keeps it from solid is control. A plain account still holds upgrade, pause and configuration power beside the 3-of-4 Safe, with no timelock, and the stake behind the validators comes almost entirely from Safes that share signers with the foundation. A token with 86% of supply still to come and a five-month public record add to that. A small USDC balance on Pharos, held to use an app, is a defensible position. The 13 to 15% vaults are a separate bet on emerging-market credit, and this report does not clear them.

Data appendix

• Scope: the Pharos Network L1 (Pacific mainnet, chain ID 1672, gas token PROS). Pharos issues no stablecoin: the docs' mainnet token registry lists Circle-deployed USDC, WPROS, LINK and WETH. Vaults distributed on Pharos (Axil Prime Credit on R25, pAlpha) are described here, not rated.
• TVL:
  • Chain TVL $74.98M on 2026-10-01 (DeFiLlama chain endpoint; DeFiLlama has no protocol slug for Pharos). That is the series high; $73.43M on 2026-09-28, $18.96M on 2026-09-01, $0.71M on 2026-07-03.
  • The chain figure adds up from Morpho Blue $69.80M, Reservoir $3.91M, FaroSwap $1.22M and Avalon $0.06M. About $47.9M of Morpho's figure is 565 BTCvc of collateral at the market oracle's BTC-parity price (on-chain), up from 138 BTCvc in early September.
  • Listed separately on DeFiLlama: R25 $134.02M (RWA vault value, high $160.27M on 2026-07-21), Ember $30.67M, BitFi Basis $28.05M, RockawayX $11.13M (curator), Chainlink CCIP $3.31M.
  • Stablecoins on Pharos $12.33M (DeFiLlama): USDC $9.13M and Multipli rwaUSDi $3.21M. Native USDC totalSupply 9,127,367 (on-chain). Axil Prime Credit vault reported assets 46,374,338 USDC (on-chain).
• Audits: github.com/PharosNetwork/pharos-audit.
  • Zellic: cryptography 2025-08-18 (1 critical, 2 high, 3 medium, 10 low; all fixed); consensus and P2P 2025-08-28 (15 critical, 2 high, 2 low; one critical partly fixed, one high acknowledged); execution 2025-09-04 (7 critical, 2 high, 3 medium, 1 low; four criticals and one high disputed by Pharos); PoS update 2025-09-30 (2 high, 1 low; one high acknowledged). 31.4 person-weeks in all.
  • ExVul: 2025-07-20 to 2025-12-01. 17 critical, 15 high, 21 medium, 34 low, 14 informational; all criticals, highs and mediums marked fixed; seven lows and one informational acknowledged.
  • OpenZeppelin system-contract re-audit: 2026-01-08 to 2026-01-16, report 2026-02-09. 27 issues (4 high), all marked resolved.
  • Node repo AntChainAldabaNG not public; PharosTumbler (consensus) and PharosCubenet (networking) public since 2026-04-15. R25: 0 audits on DeFiLlama; SlowMist per press, no public report found (refresh).
• Admin / governance:
  • DEFAULT_ADMIN_ROLE on Staking 0x4100000000000000000000000000000000000000 and ChainConfig 0x3100000000000000000000000000000000000000 is held by Safe v1.4.1 0xcce192445aa8118117e73554773e8be8617f31d0 (3 of 4, granted 2026-03-02), by the plain account 0x5f0975a92e4fb8005e7229fac96762fb3ca3d218 (no code, nonce 68, one of the Safe's signers) and by the system sender 0x1111111111111111111111111111111111111111. RuleManager 0x2100000000000000000000000000000000000000: the Safe and the system sender (plain account revoked 2026-03-02).
  • Upgrades: Staking to 0x874e39d44a2b3da5697decd9d44a2416ad8004d5 at block 5,240,286 (2026-04-14 07:30 UTC, with migrateV1ToV2) and ChainConfig to 0xd3bec4bf2c443b2dd96bec8192bb9733106d9148 at block 5,241,002 (2026-04-14 07:53 UTC), both executed by the Safe. Timelock: none found. Staking and ChainConfig paused: no.
  • Validators: 31 active, 0 pending, 115,609,950 PROS. Foundation: 4 x 5,000,000 (Safe v1.4.1, 3 of 5, 0xfa0a4299964230504081a2462f2d1478298a6549). Top 7 by stake hold 36.33%, top 16 hold 67.90%. Consensus is stake-weighted (PharosTumbler README).
  • Stake suppliers: 0x13f2a3a09c9aaa2214837aeca4801a9014e26ca7 27.0M, 0x88d0b5cabc3e2bb435c673cee534979aaceb8a70 25.6M, 0x244304c1992a1ac7c5b178d85621905130dd8170 23.5M and 0xf38d277f5c4873c981fc0cec6fbe32a0217a34f7 19.3M, each a 2-of-3 Safe with two foundation-Safe signers, plus the foundation Safe's 20.0M: 99.8% of all stake.
  • staking.withdraw_effective_epoch: 84 epochs of four hours, about 14 days (ChainConfig). Slashing: none automated (docs); slashValidator is an empty placeholder in the published source.
  • Signer identities, threshold-wallet status of the plain account, bytecode match to the audited commit, TransactionDeny address: refresh.
• Oracle: Chainlink Data Streams (verifier proxy 0xa094978891512268f4a4a4641B8da1A2a3E3BEB7), Chainlink push feeds (PROS/USD 0x371Ed36aEe5bFFbED1A2aa9b7deD7F4EB30F5466, BTC/USD and others) and Supra DORA pull (Pharos docs). Morpho Blue on Pharos: 0x18573fa18fd17ddfd790b4a5b5b2977aad3b4efb. APC3M markets (77% and 86% LLTV) priced through vault 0xd0428799fbc35557834d33121ba4472692c8908a (1 APC3M = 1.0252 USDC); BTCvc markets priced by a custom oracle at BTC parity. Cross-chain: CCTP V2 domain 31, CCIP router 0x4e52dD94e9BCfeFE3C78153bDfB0AB1d30687297, LayerZero EID 30407.
• Holder concentration:
  • PROS about $0.70, market cap about $94.9M, FDV about $699.7M, 135.6M of 1B circulating (CoinGecko, 2026-10-01).
  • Allocation: foundation 16%, Lab Co. 9%, team 20%, investors 20%, ecosystem and community 21% (6% airdrop), node and liquidity incentives 14%. Team and investors: 12-month lock, then 36 months linear. Treasury and incentive buckets: 48 to 60 months.
  • Genesis: 995,900,000 PROS to 0x00A9B001Db0409eb2a66EF1BE2f3acc5863ac405 (24,550,301 left). Live Staking inflation setting 9125 (9.125% a year) against 5% in the token plan. Top-holder table: refresh.
• Recent news scan:
  • Mainnet and TGE 2026-04-28 (The Block press release). $52M raised, including a $44M Series A; valuation reported near $1B alongside a GCL New Energy partnership.
  • R25 said its Axil Consumer Credit Vault pre-deposit campaign raised $50M and hit its cap within 48 hours (reported 2026-05-28).
  • Axil Prime Credit opened 2026-07-15: three-month lock, up to 20 days to redeem, about 14.3% gross target including PROS incentives; Binance Wallet lists 13% with a 70M USDC cap.
  • pAlpha II withdrawal requests closed 2026-10-01; deposits without a request roll into pAlpha III after 2026-10-19 (CoinGabbar).
  • Agent-Native upgrade announced September 2026.
  • No incident found on DeFiLlama hacks, web search or X search.

Maintained monthly. Methodology: DeFi Research Instruction v2.