On the evening of 29 October 2026, about $6.3 billion left Binance's hot wallets across roughly ten blockchains. The Ethereum wallet was emptied in under four minutes. It is now the largest crypto theft on record, and the FBI has attributed it to a North Korea-linked group.
The detail that stayed with me wasn't the size. It was the route in. According to Binance's own incident update, the campaign started months earlier when an engineer downloaded a tampered installer while setting up a popular AI coding tool. Later, an internal AI tool was used as a stepping stone inside the network.
AI didn't "hack" Binance. But AI tooling became part of the attack surface, and the attackers moved faster than the defenders could react.
Four failures, one pattern
Put Binance next to two other stories from the last eighteen months:
- Paraguay. In June 2025, hackers leaked records covering about 7.4 million citizens, effectively the entire population, after the government refused a ransom demand. The entry point was reportedly infostealer malware on a single government employee's device. Attacks on Paraguayan state systems have continued through 2026.
- Paraguay again, November 2026. The central bank seized Ueno, the country's largest digital bank, after its solvency ratio fell below half the legal minimum. This one wasn't a hack. But overnight, nearly 3 million customers, about half the country, were locked out of their accounts, roughly US$850 million in public money was caught up, and President Santiago Peña resigned. When trust sits in one opaque institution, its failure becomes everyone's failure.
- Microsoft. In January 2026, Microsoft 365 went down globally for roughly eight to nine hours: email, Teams and security dashboards included. In July 2026, a bug in an automated maintenance system caused another multi-hour outage across Microsoft 365 and Azure. Neither was an attack. Both showed how much of the world's work runs through one company's shared infrastructure.
A crypto exchange, a national government, a digital bank and the world's biggest software company. Different industries, same weakness: a central point that, once compromised or broken, takes everything down with it.
Why AI changes the math
Single points of failure aren't new. What's new is the speed and scale of the people probing them.
TrendAI's 2026 survey of financial-sector security chiefs found 89% saw a rise in AI-enabled attacks, with criminals chaining specialised AI agents to run reconnaissance, phishing and exploitation at machine speed. BioCatch's 2026 survey of more than 1,400 fraud leaders found 80% had already faced attacks using agentic AI.
When attackers can run thousands of patient, automated attempts in parallel, every central database, shared login system and hot wallet becomes a bigger target. Defending a single point gets harder every year.
What tamper-evident, programmable ledgers actually offer
Let me be precise, because this is where people overclaim.
A shared ledger does not make money unhackable. Binance is proof: by most analyses the blockchains worked as designed; the weakness was the custodian's internal operations.
What a well-designed ledger does offer is three things that matter more in an AI-speed world:
- Tampering is visible. Every movement is recorded and auditable in near real time. Within hours of the Binance theft, analysts were tracing funds across chains.
- No single off switch. Records are replicated, not held in one database that one bug, or one stolen password, can take down.
- Rules can be coded into the money itself. Spending limits, time locks, multi-party approval and freezes. After Binance, Tether froze more than $1.5 billion in USDT linked to the theft, and the Arbitrum Security Council forced about $30 million out of an attacker's wallet.
That last point cuts both ways. Programmable money can be frozen to stop a thief, and it can also be frozen to control a citizen. Which is exactly why the next decade's fight is about who writes the rules.
Two roads to the same destination
Governments are already splitting into two camps.
Sovereign digital currencies. China's retail e-CNY had processed more than 3.4 billion transactions worth about 16.7 trillion yuan by December 2025. India is piloting the digital rupee. The ECB says a digital euro could be issued in 2029 if EU lawmakers pass the regulation in 2026.
Regulated stablecoins. The US has gone the other way. A temporary ban on a Federal Reserve digital currency until 2030 became law in July 2026, and policy is pushing digital-dollar use towards regulated, fully backed stablecoins.
They look like opposites, and politically they are. But technically both are the same move: money that lives on programmable, auditable ledgers instead of in a patchwork of private databases. The argument is over who controls the ledger, not whether the ledger wins.
The honest caveats
- CBDCs raise real privacy and surveillance concerns. Europe is still arguing about holding limits and privacy thresholds; many US lawmakers oppose a CBDC on exactly these grounds.
- Adoption is slow. Most CBDCs are still pilots.
- Custodians, wallets and bridges remain weak points. The ledger is only as safe as the keys and operations around it.
So my claim isn't "blockchain will stop hacks". It's narrower: as AI makes attacks faster and cheaper, the financial system will move towards ledgers where tampering is visible, failure isn't centralised and safety rules are built in, whether they're issued by central banks or regulated private firms.
Why this matters beyond finance
I work in advertising, and this matters to us more than it seems.
Programmable money is what makes AI agents economically useful. An agent that can pay, get paid and prove what it was authorised to spend opens up new models: micro-payments for content, rewards paid instantly, value exchanged with consumers for attention or data with clear consent, and agent-to-agent commerce that never touches a checkout page.
The brands and platforms that understand programmable money early will design the next generation of commerce and loyalty. Everyone else will be integrating it later, on someone else's terms.
— Aditya Tibrewal
Sources
- Binance hack timeline and scale: https://socradar.io/blog/binance-hack-6-3b-hot-wallet-theft/
- Binance incident update (AI coding tool installer, internal AI tool): https://www.binance.com/square/post/384314395829022
- FBI attribution: https://www.financemagnates.com/trending/binance-hack-began-with-fake-job-interviews-ended-with-63-billion-stolen/
- Tether freeze: https://bleepingcomputer.com/news/security/binance-probes-potential-58-billion-crypto-theft-hack
- Paraguay 7.4M records leak: https://securityaffairs.com/178970/data-breach/paraguay-suffered-data-breach-7-4-million-citizen-records-leaked-on-dark-web.html
- Ueno digital bank collapse, Paraguay (Nov 2026): https://www.retailbankerinternational.com/news/ueno-bank-collapse-exposes-oversight-gaps-digital-banking-globaldata/
- Ueno public funds and president's resignation: https://www.nst.com.my/amp/opinion/columnists/2026/11/1573743/how-digital-banks-collapse-toppled-paraguays-president
- Paraguay infostealer entry point: https://blog.rankiteo.com/pre301062425-government-of-paraguay-cyber-attack-june-2025/
- Microsoft 365 outage, Jan 2026: https://www.webpronews.com/microsoft-365-outage-halts-global-operations-on-jan-22-2026
- Microsoft maintenance-bug outage, Jul 2026: https://www.messageware.com/microsoft-365-down-a-maintenance-bug-not-an-attack/
- TrendAI Modern Bank Heists 2026: https://www.techcentral.ie/?p=156866
- BioCatch 2026 survey: https://www.technologydecisions.com.au/content/security/news/banking-security-leaders-see-ai-as-the-industry-s-top-threat-1370163798
- CBDC status 2026: https://eco.com/support/en/articles/15210561
- US CBDC ban to 2030: https://forklog.com/en/u-s-senate-passes-housing-bill-with-ban-on-a-fed-cbdc-until-2030/