Authorized is not verified: what the Binance breach shows for tokenised assets

On 29 October 2026, Binance reported a theft of about $6.3 billion from its hot wallets. NBC News reported the company statement the same day. Binance said a compromised DevOps engineer's laptop let an attacker reach a service container and submit unauthorised withdrawals. It said it had not found evidence of a leaked private key. User funds would be covered, it said, and deposits and withdrawals were suspended.
Elliptic, in analysis last updated on 1 November, puts the theft at $6.43 billion across 11 blockchains, more than four times the $1.46 billion taken from Bybit in February 2025. Binance's figure remains about $6.3 billion. Both numbers are on the record. Elliptic also records that Binance later said the investigation had found additional points of breach, and that Binance co-CEO Richard Teng described a complex attack that was not carried out by one person.
The useful lesson is not that an exchange was robbed. It is that authorized is not verified.
On Ethereum, Elliptic and CoinDesk describe the same path. EIP-7702, introduced in the 2025 Pectra upgrade, lets an ordinary account delegate its behaviour to a smart contract. At 20:20 UTC on 29 October, a single transaction carried 50 authorisations, each signed with the key of the Binance hot wallet known as Binance 14, each covering a different nonce. One matched. Control of that wallet was delegated to a sweeper contract. Minutes later the wallet was emptied of 167,464 ETH and tokens that included about 1.6 billion USDT. The delegation stayed in place, so later deposits could be swept without a new signature. Teng said the EIP-7702 path did not pass through Binance's normal approval controls.
Elliptic records that, on the other affected chains, funds left Binance wallets in ordinary signed transactions. On Ethereum, a second hot wallet, Binance 16, also sent assets worth about $174 million directly to the attacker, signed with its own key. The evidence Elliptic sets out is that the attacker could instruct Binance's signing systems, rather than steal a key and use it outside those systems.
On 31 October, the Federal Bureau of Investigation attributed the theft to North Korea, naming the activity "TraderTraitor". The FBI said more than $6 billion in virtual assets was taken on or about 29 October, and that the actors were converting stolen assets primarily into ether. Elliptic had assessed TraderTraitor as the likely perpetrator before that announcement.
Issuer freezes then showed what a token can do that a native asset cannot. Reuters, as carried by The Hindu Business Line, reported that Tether froze nearly $1.57 billion and Circle froze more than $800 million, together more than $2.38 billion. Ripple froze roughly $200 million, Binance co-founder Yi He said. Elliptic's 1 November update verifies $2.85 billion of frozen or seized funds on-chain, close to Binance's estimate of about $3 billion. Almost all of that frozen stock was stablecoins. Ether and bitcoin have no issuer that can blacklist an address. Elliptic records that nothing stolen on bitcoin, the XRP Ledger, Solana, or NEAR has been frozen for that reason, and that the attacker's remaining unfreezable stock stood at about $3.25 billion.
A freeze is also not a recovery. A blacklisted address cannot move the frozen token. That does not return the asset to the victim. Reissuance, if it comes, is a separate act by the issuer. Containment and restitution are not the same step. On 1 November, Reuters reported that Binance had begun a phased resumption of withdrawals and said it had covered about 95 percent of the affected assets from company funds. That cover is a balance-sheet decision by the exchange. It is not the same as recovering the stolen assets from the attacker.
For tokenised securities, the instinctive reading will be that permissioned assets are safer because transfers can be restricted and addresses can be frozen. That reading is incomplete. The Binance path did not begin with a broken contract. It began with systems and people that could produce a signature the network would accept. The instruction was authorized. It was not verified as a legitimate user action.
That is the structural point. An authorisation layer that accepts a single valid signature as proof of a legitimate instruction will do what it is told. For an exchange hot wallet, the loss is a liquid balance. For a tokenised claim on an energy asset or a revenue stream, a transfer that the ledger treats as valid can still leave the legal claim contested for years. The record shows a completed move. It does not show that the move was intended.
At Juggervault Finance, the offering documents carry the claim, and the ledger is a record of that claim, not the right itself. Creating project shares requires a typed platform signature that is bound to one specific offering, cannot be reused, and is issued only after a compliance approval is recorded.
For an institution evaluating a tokenised asset platform, three questions follow directly from the Binance record.
- Does the platform require two independent approvals, not two signatures from the same key, but two separately secured parties, before an asset moves?
- Is each approval bound to a specific action, such as creating project shares, opening a primary sale, or resetting a buyback price, or is it a general credential that can be reused?
- Can a single compromised system or stolen credential produce a valid instruction on its own, or does the architecture require a second party that cannot be forged from the first?
A platform that can answer those questions in the affirmative has separated authorization from verification. A platform that cannot has not.
The question for an institution that holds, issues, or settles a tokenised asset is whether its own approval path would treat an authorized instruction as a verified one. If it would, the Binance record is the case for changing that path.