BounceBit kills its own L1 after $3M Evmos bug, and reissues BB on BNB Chain

The bug was not in code BounceBit wrote. According to the company, the failing authorization check was part of a built-in feature of the Evmos stack that BounceBit Chain was forked from, and it let a caller nominate any account as the source of a transfer without that account ever approving it ^[4]. That is why BounceBit can say, accurately, that no private keys were compromised and no signatures forged ^[5]. The signature layer worked as designed and it made no difference: nine mainnet accounts were emptied by a caller who never needed their keys ^[1].

Containment worked too. The incident stayed inside the chain, and the CeDeFi Strategy, Promo Vaults, Prime and real-world asset products kept running ^[6]. Both statements are true and the chain is still being switched off, which is the part worth sitting with.

The reason is upstream. Evmos itself was discontinued earlier in 2026 ^[9], so there was no maintainer to publish a fix and no supported branch to track. The choice reduced to operating an unpatched chain or operating none.

Some arithmetic the announcement leaves to the reader. The exploit window ran 4 hours and 52 minutes ^[17], and 5 hours and 34 minutes passed between the first unauthorized transfer and the halt of block production at 02:36 UTC ^[18]. The $3 million valuation implies roughly $0.0105 per BB ^[19], so this was a cheap theft in dollar terms and an expensive one in every other. The reissue snapshot is taken at block 20,697,260, while the chain stopped at 20,702,857: 5,597 blocks that still exist as blocks but no longer exist as balances ^[20].

That gap is the actual remedy. The ~286.5 million moved tokens are excluded from the new supply ^[12], legitimate holders including staked and unbonding positions receive BEP-20 BB automatically at matching addresses with no claims process ^[13], and exchange balances are being adjusted by coordination rather than by consensus ^[14]. A rollback of this kind is normally the hardest thing a network can attempt. It is straightforward here because there is no longer a network to object.

BounceBit also frames the move as product logic: a standalone Layer 1 no longer matches user needs, and most of its products and activity already sit on BNB Chain ^[10], where it cites more mature security infrastructure, deeper liquidity and wider wallet compatibility ^[11]. Read against the dates, the two rationales do different work. The migration may well have been coming; the timing was set by a chain nobody could patch.

For anyone whose pitch is restaking, the lesson is about where collateral sits. A forked L1 is not a one-time engineering cost, it is a permanent maintenance liability whose worst case is the upstream project ending while your users' assets are still on your copy of it. BounceBit's own products were fine throughout ^[6]. The venue was not, and the venue was the thing users were told to trust.