The attack and its aftermath: On August 12, 2026, Harmony Protocol detected a major exploit where attackers were able to mint unauthorized ONE tokens. It turns out there was a cross-shard receipt replay vulnerability that basically let them process the same receipt over and over again, thus creating tokens without corresponding debits.
At first, it looked like 4 billion ONE were minted in a couple of empty blocks, but once the full picture was reconstructed, it was revealed that a massive 3.01 trillion tokens were forged across six cross-shard transactions to four exploiter wallets. One wallet attempted 534 transfers of 5 billion ONE each in 106 seconds; 477 succeeded, moving 2.385 trillion tokens.
Why the Harmony Protocol’s rollback was chosen
Per their communication, the Harmony blockchain team evaluated several alternatives before settling on a fixed rollback window. A targeted burn was deemed unsafe because forged tokens had already moved through exchanges, decentralized exchange pools, bridges, and liquidity positions; destroying them could burn innocent users’ funds. For instance, a blacklist wouldn’t remove the forged supply and could block unrelated wallets.
Trying to redo specific transactions was a no-go, since a state change meant the same transaction might produce totally different results. And trying to migrate all those tokens would have just caused even wider disruption.
The team concluded a single network-wide rule was “the fairest and most secure” approach.
What happens to users
Validators will receive replacement databases for both shards and resume block production from heights 92,730,035 and 94,978,279. All blocks after the checkpoints will be discarded, including legitimate transactions.
Harmony analyzed 109,126 regular transactions in the affected window; 95.8 percent were automated, mostly DEX operations. The team noted that “of the options studied, one fixed rollback window is the fairest and most secure.”
Validators face a technical tightrope as rollback execution begins
The success of this rollback hinges entirely on Harmony’s validator network, which must stay perfectly in sync to make sure the chain doesn’t split. Validators need to download replacement databases for both shards, verify block hashes, and await Harmony’s official “GO” signal before restarting production.
A handful of validators, like kratos_harmony, Fortune_ONE, and LegionValidator, are already getting ahead of things. They have been testing recovery scripts on all sorts of hardware setups to iron out any bugs before the main rollout happens.
The coordination window is super tight. If a validator is running outdated software or misses the window for the migration, they could end up on the wrong fork, which would just fragment the network even more.
Harmony Protocol’s recovery guide is pretty clear: don’t restart until you get the green light. It’s a high-stakes move, so they are trying to balance moving fast with actually staying safe.
