sBOLD Vault Exploit: Share-Accounting Flaw Drained About $3.4 Million

A November 19, 2026 report by Deep Tide TechFlow said an attacker drained about $3.4 million in liquidity from K3 Capital’s Ethereum sBOLD savings vault. The report described 97 transactions over roughly eight minutes and attributed the exploit to how the vault accounted for the value of assets transferred into it.

How the reported exploit worked

sBOLD is an interest-bearing vault built around BOLD, the stablecoin associated with Liquity. According to Bitquery’s transaction analysis, the attacker repeatedly deposited BOLD, introduced wrapped ether (WETH) that was not reflected in the vault’s balance-sheet accounting, and converted it internally to BOLD. This inflated the redemption value of each sBOLD share, allowing withdrawals against other depositors’ funds.

MEV competition and proceeds

Bitquery reported that an automated bot detected the original attacker’s pending withdrawal transaction, copied and changed its parameters, and extracted about $1.91 million—56.18% of the reported $3.4 million total. The bot paid roughly $1.78 million in fees to block builders to secure transaction ordering, leaving about $130,000 before any other costs. The report said Titan handled most blocks containing the bot’s transactions, while BuilderNet, Ultrasound and Quasar also participated in ordering relevant blocks.

The $1.78 million was gross revenue to builders, not necessarily their profit: builders still had to bid for blocks through MEV-Boost. The report put the original attacker’s and initial funding wallet’s combined net proceeds at about $520,000.

Earlier warning and unresolved questions

ChainSecurity audited sBOLD in May 2025 and warned of a route to extract excess value through self-liquidation. Whether this attack used that route remained under investigation at the time of the November 19 report. The report also said the team was tracking funds, assessing depositor losses and working on a vault-code fix.

For ERC-4626-standard vaults generally, the report identified restrictions on unapproved token transfers, per-block redemption limits and time-weighted oracles as potential safeguards. It did not establish that those measures had been deployed for sBOLD.

Further reading