On 18 November 2026, Gonka's Ethereum bridge was used to release stablecoin reserves and to mint Wrapped Gonka, or WGNK, that Blockaid said had no corresponding backing. Gonka is a network for decentralized AI compute. Its native coin, GNK, moves to Ethereum through this bridge. Gonka's documentation says a single Ethereum contract, 0x972a7A92D92796a98801A8818bcF91f1648f2F68, is both the bridge and the WGNK token.

At 18:00:59 UTC, Etherscan shows that wallet 0x6650a1CB0c3B970a0e4CEd26984fe06DB7844D80, which Blockaid named as the suspected exploiter, called submitGroupKey on that contract. The call succeeded and a GroupKeySubmitted event was emitted. Blockaid said an attacker had rotated the bridge's epoch BLS group key and used it to empty remaining stablecoin reserves and mint unbacked WGNK. That describes the sequence on chain. It does not establish how the caller obtained a signature the contract would accept, or why the update was valid. The root cause is unconfirmed.

How the funds moved

The Defiant, citing Ethereum records, reported withdrawals of 884,151 USDT and 648 USDC, plus the creation of 22 million WGNK. On-chain breakdowns reported by The Crypto Times, citing the investigator TraceFast, and by LC Insight, citing Etherscan, give the same transfers as 884,150.87 USDT and 647.74 USDC at 18:01 UTC, and 22 million WGNK minted about a minute later. Onchain Lens put the stablecoin withdrawals at about $884,800 and valued the newly minted WGNK at about $172,000, for a total of about $1.06 million in affected assets.

Blockaid, posting at 18:28 UTC, estimated about $1.05 million realized from the bridge drain across roughly five transactions and said the attack was still ongoing. That figure is a preliminary realized-value estimate at the time of the alert, not a reconciled final loss, and it should not be added to the stablecoin withdrawals or to the 22 million token mint. The mint is a quantity of WGNK, not a second dollar amount. Researchers later reported that the withdrawn USDT was swapped through Uniswap and that ether moved on to a second wallet. LC Insight, reading Etherscan, put those proceeds at about 606.7 ETH.

Gonka's response, and the ether that came back

At 00:18 UTC on 19 November, Gonka's official account said the incident had been detected within minutes and that core contributors were investigating, but that details could not yet be shared. The post asked those behind the incident to return 100 percent of the assets taken within 12 hours, and published return addresses on Gonka and on Ethereum, the latter being 0x1d806D0E3C7Df74C14A49d58De98c189614E1768.

A second official post, six minutes later, said the incident did not affect network consensus and that inference services continued to operate normally. Until further notice, Gonka told users not to use or trade WGNK or the bridge, and to withdraw WGNK from liquidity pools on Uniswap and other exchanges. Whether WGNK holders should be compensated, it said, is a question for the community. Gonka has not published a postmortem, and it has not confirmed how the group key was changed.

At 07:40:35 UTC on 19 November, Etherscan shows that Gonka's published Ethereum return address, a Safe, received 593.435019683433344968 ETH in a single transfer. The sending wallet, 0xb916f815DD02aEa48957F3679a7534DbD2238060, had been funded minutes earlier by THORChain's router. At 09:13 and 09:14 UTC, the official account wrote that the funds had been returned, and then that they had been fully returned.

The 593.44 ETH credit is independently visible on Etherscan. It is close to Blockaid's preliminary $1.05 million estimate and to the ether proceeds reported from the earlier swaps, but it is not a match to the last decimal, and this review has not traced the THORChain funding back to the exploiter wallets. A return of ether also does not, by itself, retract the 22 million WGNK or restore one-to-one backing. Gonka has not followed the short return posts with a statement that the bridge is safe to use again, or that WGNK is fully backed.

What this leaves open

A bridge that will honor a new group key will honor whatever that key then signs. The public record shows a successful key submission from a previously unused wallet, then withdrawals and a large unbacked mint in roughly a minute. It does not yet show why that submission was accepted. Until Gonka publishes a root-cause analysis, claims of a stolen key, a forged signature, or a specific contract bug should be treated as unconfirmed.

The rest of the defensive work is specific. It includes how a bridge should authenticate an epoch key rotation, what independent controls should sit in front of mint and withdraw, how quickly a bridge can be paused once a key change is observed, and how wrapped-token holders are treated when backing is broken. Those checks, and a fuller technical walkthrough of this incident, are part of the Advanced Security Academy. The waitlist is coming at advancedsecurity.club.