Mentioned in This Article
1. Why is Changpeng Zhao warning about crypto hacks and AI?
Changpeng Zhao, the founder of Binance, warned on November 3 that artificial intelligence could drive a new wave of crypto hacks aimed at exchanges and banks, following the $6.4 billion theft from Binance on October 29. He said the industry has to use AI to stop AI-driven attacks, adding that there is no other way.
Zhao, who no longer runs the exchange, made the remarks in an exclusive interview. His central argument was one of speed: in his view, hackers are gaining access to AI capabilities faster than financial institutions can put similar tools to work in their defenses. He gave no evidence that AI played a part in the Binance breach itself, and his comments should be read as an opinion and a forecast rather than a finding about how the theft was carried out.
2. Why do attackers hold the advantage?
Zhao framed the problem as an imbalance. An attacker, he said, needs to find only one crack, while a defender has to find and close all of them. That asymmetry is a familiar principle in security, and it is the reason a single overlooked weakness can undo years of careful engineering.
His concern is that AI tools reduce the cost of searching for such weaknesses. Automated systems can examine large volumes of code, configurations and employee behavior far faster than a human team, and the same capability is available to the people trying to defend a system. The question he raised is which side adopts it more quickly. According to Zhao, the attackers are currently ahead of the institutions they target.
3. What do investigators know about the heist?
The warning follows the October 29 theft from Binance. The FBI attributed the heist to TraderTraitor, a hacking group associated with North Korea. Zhao did not mention North Korea in his remarks and concentrated instead on the security risks that AI poses to exchanges and banks.
The reported size of the loss differs slightly between accounts. The figure used for the theft in this interview is $6.4 billion, while Binance's own briefing the same day described a $6.3 billion hack. The difference was not explained, and both numbers refer to the same incident.
4. Did AI launder the stolen Binance funds?
Investigators found no evidence that artificial intelligence was used to launder the stolen money. Elliptic, a blockchain analytics firm, said the transactions appeared to follow automated scripts rather than decisions made by AI agents.
Tom Robinson, a co-founder of Elliptic, said the firm's assessment is that the laundering is automated, but that nothing it has seen shows an AI agent making decisions. The distinction is important. A script follows fixed instructions written in advance, such as splitting funds across many addresses on a schedule. An AI agent would choose its own actions in response to events. Elliptic's observations point to the first of these.
Those findings also have a limit. They describe how the stolen funds moved after the theft, not how the attackers first got into Binance. For that reason, Elliptic's work neither supports nor contradicts the view that AI helped with the initial breach, and Zhao offered no evidence on that point either.
5. How could automation lower the barrier to attacks?
The phrase at the center of Zhao's warning is that automated tools lower the barrier for complex breaches. Complex intrusions have traditionally required scarce expertise: people who can write convincing messages, find vulnerabilities, adapt malicious code to a particular environment and cover their tracks. Each of those tasks takes time and skill, which limits how many targets a group can pursue at once.
AI software can shorten several of these tasks. It can draft messages in fluent language, summarize large amounts of technical material and suggest ways to modify code. If such tools become cheap and widely available, groups with modest resources could attempt operations that were once reserved for the best-funded teams, and well-funded groups could run more of them in parallel. That is the mechanism Zhao has in mind when he predicts more hacks in the short term.
It is worth separating this general argument from the facts of the Binance case. The evidence reported so far on the laundering points to ordinary scripts, and Zhao supplied nothing tying AI to the intrusion. His warning concerns the direction of travel for the industry, not a documented cause of this particular theft.
6. What would AI-based defenses look like?
Zhao did not lay out a technical design, but his statement that the industry must use AI to prevent AI hacks points to a few obvious categories. Automated tools can scan an organization's own code and systems for weaknesses before attackers find them, monitor employee accounts for unusual behavior, and flag transactions that depart from normal patterns. In each case the aim is to give defenders the same speed that attackers are expected to gain.
The difficulty, as Zhao described it, is the pace of adoption. Financial institutions are generally cautious about deploying new software that touches sensitive systems, and they face compliance reviews and internal approvals that attackers do not. That gap between caution and speed is what he said hackers are exploiting.
His call for common standards fits the same picture. If firms agree on baseline practices and share what they learn about attacks, the burden of finding every crack no longer rests on each company alone. Whether the discussions he mentioned lead to a formal alliance is not yet known.
7. Are banks at risk too, and what does Zhao propose?
Zhao said the danger reaches beyond crypto. In his opinion, if an exchange can be hacked, banks are in a worse position. His reasoning was that traditional institutions could face comparable threats as AI improves what attackers are able to do.
This was presented as his view, not as a conclusion drawn from the Binance investigation. It is nonetheless consistent with his broader message: the tools that make complex intrusions easier are not specific to digital assets, and any organization that holds valuable data or money is a possible target. Banks, exchanges and payment firms all depend on software, credentials and employees that can be probed in similar ways.
Zhao called for common security standards or an industry alliance that would help firms respond to attacks. He said discussions about such coordination were underway, though nothing had been agreed and no details were provided.
The idea reflects a recurring theme after large breaches. Individual companies often learn about attacker techniques only after they are hit, whereas a shared standard or alliance could allow lessons to spread before the next victim is chosen. Whether competitors in the exchange business can agree on such an arrangement remains an open matter, and the interview did not suggest a timetable.
8. How have customers reacted at Binance?
The interview also touched on the movement of customer money after the theft. Zhao said an initial $3.6 billion in outflows represented about 2% of Binance's holdings, which implies assets of roughly $180 billion. He had expected withdrawals of 10% to 30% of that amount, or between $18 billion and $54 billion, so the actual early outflows were far below his own expectation.
Binance chief executive Richard Teng shared separate figures covering a 46-hour period that ended on Tuesday. They showed $10.97 billion in withdrawals and $5.28 billion in net outflows. Teng said hourly deposits had started to exceed withdrawals, although the exchange was still in net outflow over the full period. Net outflow means that more money left than arrived, and it is the number that best reflects the pressure on the exchange's reserves.
9. How much of the stolen money has been frozen?
Zhao said that Tether, Circle and law enforcement agencies helped freeze stolen funds, with the U.S. Secret Service assisting. Tether and Circle are the issuers of the USDT and USDC stablecoins, and issuers of this kind are able to block tokens at particular addresses when asked to by authorities.
Researchers estimated that $2.85 billion of the stolen funds had been frozen or seized. That is a large share of the total, though it still leaves a substantial amount outside the reach of those freezes, and the interview did not discuss where the remaining funds are.
The role of stablecoin issuers illustrates a feature that sets some digital assets apart from cash-like tokens with no central party. Because Tether and Circle control the contracts behind their dollar-pegged tokens, they can respond to a law enforcement request by blocking movement at identified addresses. That ability depends on speed, since funds that have already been swapped into other assets or moved through many addresses are harder to stop. Zhao's account suggests that cooperation between the exchange, the issuers and agencies such as the U.S. Secret Service was a central part of the response in the days after the theft.
10. What happens next?
Zhao expects the next period to be difficult. In his words, the industry is going to see a lot more hacks in the short term. Over a longer horizon, he said, AI could strengthen financial security and reduce losses, provided that defenders use it as quickly as attackers do.
The points to watch are whether the industry alliance he described takes shape, whether further analysis shows any role for AI in the Binance breach, and how the exchange's flows develop if hourly deposits continue to exceed withdrawals. None of these has a confirmed date. For now, the available evidence shows scripted laundering and a human-led intrusion, while the AI threat remains, in Zhao's account, a warning about what is coming.