A single compromised laptop recently led to a $6.3 billion cryptocurrency theft, reportedly the largest crypto heist in history by dollar value. In another attack, a vulnerability in Bitcoin wallet software allowed hackers to steal approximately $120 million from roughly 5,000 offline storage devices.
As I have covered in previous essays and episodes, artificial intelligence threatens to make attacks like these much easier to carry out. It can find security weaknesses in minutes that once required time and substantial expertise. And the same capabilities could be directed at banks, hospitals, government agencies, nonprofits, and virtually anyone else connected to the internet.
Rob Hamilton, cofounder and CEO of AnchorWatch, a Bitcoin custody and insurance company, has been fighting these attacks. He expects a dangerous year or two ahead, but believes AI could ultimately make the internet much safer. His experience helps explain why both predictions may be right.
How One Laptop Exposed Billions
On October 29, Hamilton followed the Binance attack as stolen cryptocurrency moved across multiple blockchains in approximately twenty minutes.
The attackers apparently compromised a developer’s laptop, gaining access to software controlling enormous amounts of money. They did not need to steal the private keys or break any blockchain’s cryptography; they exploited human and organizational weaknesses in the systems controlling access to the assets.
“If any one engineer can deploy code that touches six billion dollars, that is the failure,” Hamilton told me.
Hamilton believes AI may have helped the attackers map Binance’s systems and identify their target. But the decisive compromise involved a person with extraordinary privileges. Why was so much money exposed to a single point of failure? Dividing assets among independently secured wallets, with limits on withdrawals and software changes, could have substantially reduced the damage.
Why Security Alone Isn’t Enough
Despite the extraordinary theft, the broader financial consequences were limited. Binance committed to covering customer losses, and withdrawals resumed after a brief disruption. Tether and Circle helped freeze approximately half the stolen funds, limiting the attackers’ ability to move them.
I asked Hamilton whether an insurance company evaluating Binance’s risk would have discovered the weakness. “Yes. I would say so,” he replied. An insurer would have examined who could modify systems controlling billions of dollars and what safeguards existed against catastrophic failure, and would likely have found the weakness. Hamilton also doubted Binance could have insured more than a small fraction of such enormous exposure.
Security is about preventing an attack, while resilience is about limiting the damage when one succeeds. Hamilton believes both are essential because no system can be perfectly secure.
AnchorWatch, the company Hamilton founded, is built around that principle. Its custody system is designed so that no single compromised laptop or employee could move customer funds: the company acts as a cosigner but can never move customers’ Bitcoin on its own, and its signing keys remain offline. Insurance then covers what security cannot prevent, with policies that name each customer directly. Most Bitcoin holdings carry no such protection, and customers of a failed custodian typically become unsecured creditors.
25 Volunteers, 10,000 Potential Vulnerabilities
The Coldcard wallet attack exposed a different kind of weakness. Its software generated insufficiently random secrets, leaving supposedly secure Bitcoin vulnerable. The flaw had gone undetected for years, until, Hamilton believes, AI made it possible to discover and exploit it.
Hamilton believes the attacker was “pretty obviously not sophisticated,” judging by how the stolen Bitcoin was moved. After the incident became public, Hamilton used an open-weight AI model to find the same flaw for about one dollar, roughly what it would have cost anyone who thought to look.
He then helped organize the Bitcoin Red Team. Approximately 25 volunteer engineers began examining other Bitcoin software for similar weaknesses, identifying roughly 10,000 potential vulnerabilities and contacting developers to get confirmed problems addressed.
The entire undertaking cost approximately $20,000 to $25,000, Hamilton told me, and was later reimbursed by a nonprofit organization. The effort was organized entirely by volunteers, without the backing of a large corporation, government agency, or other major institution.
The Open-Source Trade-Off
The Red Team relied on open-source AI models to uncover security vulnerabilities. When Hamilton turned to more powerful models from OpenAI and Anthropic to help fix confirmed flaws, their safety restrictions sometimes prevented him from doing so, though he says those restrictions have since eased somewhat. Legitimate security researchers can look remarkably like hackers, and AI companies have too much at stake to risk helping criminals by mistake.
Hamilton’s experience undercuts growing calls to restrict or ban open-source AI models. It is true that anyone, including criminals, can download and use these models without meaningful oversight. But that same freedom allows independent researchers and volunteers to defend systems without permission from large technology companies.
Restricting open-source AI could therefore weaken defenders while doing little to stop criminals who will seek unrestricted tools anyway. The Red Team’s experience illustrates what could be lost: the ability of independent researchers to use powerful AI to protect the systems we all depend on.
The Dangerous Road to a Safer Internet
Hamilton expects the next year or two to be “incredibly volatile” as criminals exploit weaknesses faster than organizations can repair them. Autonomous AI agents could eventually search enormous amounts of software with little human supervision, and criminals may adopt these tools more quickly than defenders.
Defenders have a clear advantage in numbers: banks, technology companies, governments, and cybersecurity firms employ far more people protecting digital systems than criminals employ attacking them. But those resources are spread across thousands of organizations, each guarding its own systems, while attackers can concentrate on whichever target looks weakest. AI sharpens that imbalance because, as Hamilton put it, these tools are “massive force multipliers for a single person being able to do a lot.”
Yet once a vulnerability is identified and corrected, that particular avenue of attack can be closed. AI can help defenders systematically eliminate weaknesses that might otherwise have remained undiscovered for years.
Cryptocurrency may be an early warning system for the wider economy, since stolen Bitcoin is hard to recover and attacks there become visible immediately. Hamilton worries about banks, hospitals, utilities, and other institutions that may be less prepared for AI-enabled attacks. Some may already be experiencing breaches that never become public.
He describes the opportunity ahead as removing ‘30 years of human tech debt’: decades of software weaknesses accumulated as people built increasingly complex systems. AI could allow defenders to find and repair those weaknesses at a scale previously unimaginable, but no system will ever be perfectly secure. A safer internet will also depend on resilience, with systems built so that a single compromised laptop can never again put billions at risk.
Hamilton expects serious disruption before that happens, and he believes the attacks will make institutions rethink practices they have tolerated for years. As he put it, a lot of custody has been run “a little bit too casually for too long.” If he is right, the attacks now exposing thirty years of tech debt are also the start of paying it down.
Full interview.
Thank you very much to our sponsors:
Foundation: When it comes to protecting your Bitcoin, you shouldn’t have to simply trust the company that made your hardware wallet.
Foundation believes in verifiability over reputation. Its firmware and apps are open source, its builds are reproducible, and both Passport Core and Passport Prime have undergone independent hardware security audits. The point isn’t simply that Foundation says its products are secure. It’s that you can verify that for yourself.
Foundation is also expanding what users can control:
Passport Prime goes beyond Bitcoin. It can manage passwords and two-factor authentication codes, serve as a FIDO2 security key, and provide encrypted storage—all on a device you own and control. Foundation also plans to enable app sideloading, opening Prime to applications from independent developers.
Users can now add their own entropy. The latest Envoy and KeyOS releases allow users to introduce their own randomness when creating a seed, reducing reliance on the device itself to generate it.
Order a new Passport Prime and receive a free bumper case using Code PROSPERITY at → foundation.xyz
River Financial: If not cold storage for your Bitcoin, I’d highly recommend River for buying and holding- River is engineered to protect your Bitcoin over the long run. Use this link for discounts: https://river.com/signup?r=T7GGAF7G