One Laptop, Seven Keys: Halborn and the $36M Humanity Hack

The Hack That Proved Its Own Point

On 9 June 2026, a protocol whose entire reason for existing is to prove that you are a real, unique human being lost roughly $36 million to the most ordinary human mistake in security: someone kept the keys in the wrong place. Humanity Protocol markets palm scans and zero-knowledge proofs, a privacy-preserving way to show that you are a person and not a bot. None of that cryptography mattered on the day it was drained. The trust root of the whole system turned out to be a single developer laptop, and that laptop was infected.

CoinDesk put the lesson in its headline: the exploit happened because a multisig wallet effectively lived on one laptop. The arrangement meant to stop any one person, or any one compromised machine, from moving funds had enough of its keys sitting, backed up, on a single box that one attacker could satisfy the signing threshold alone. The sophisticated part of the stack was never touched. The boring part failed completely.

The timing made the point twice. As this piece was being written, the industry was still counting the damage from the largest crypto theft ever recorded: on 29 October, about $6.36 billion walked out of Binance in roughly twenty minutes, an attack that Binance chief executive Richard Teng attributed to an abused EIP-7702 delegation on a hot wallet rather than to any flaw in a token contract. Different scale, same shape. The code was not the weak point. The signing was.

The firm that wrote the definitive post-mortem on the Humanity breach was Halborn, the Miami security company that has quietly turned the exploit autopsy into a product. Halborn did not build Humanity Protocol, and there is no sign it audited the contracts. It did what it increasingly does: read the body and publish the cause of death. The cause, once again, was keys.

The Firm That Reads the Body

Halborn was founded in 2019 by Steven Walbroehl and Rob Behnke, and it spent its early years as an offensive-first shop, the people a project hires to break in before a criminal does. It bootstrapped for roughly three years before raising a $90 million Series A in July 2022 led by Summit Partners, and it has since expanded from pure smart-contract review into red teaming, infrastructure security, and incident response. Under chief executive Jacques Boschung, who joined in 2024 from the traditional cybersecurity world, the firm has pushed hard into banks and tokenized assets; its own site now says it works with six global systemically important banks alongside many of the largest protocols in crypto, under a pared-down mission to “enable and secure the world’s digital asset adoption.”

One detail on that services page matters more than the rest for this story. Sitting right next to the smart-contract audit is a line called the Custody and Key Management Assessment. That product exists because of incidents exactly like Humanity. A code audit reads Solidity; a key-management review asks a different and less comfortable question: where do your keys actually live, who can touch them, and what happens when one machine is owned?

Halborn’s public Explained series has become a running casebook of how crypto broke in 2026, and the pattern in it is impossible to miss. Across the year’s marquee losses, the contracts almost always worked exactly as written. What failed was everything around them.

A Protocol Built to Prove You Are Human

Humanity Protocol is a biometric identity network. Its pitch is a privacy-preserving proof of humanity built from palm-print and vein-mapping biometrics, zero-knowledge proofs, decentralized identifiers and verifiable credentials, combined to let a person prove they are real and unique without handing over the underlying data. It is one of the higher-profile answers to the same problem Worldcoin tried to solve with iris scans, led by founder Terence Kwok. The H token is the network’s native asset.

The irony writes itself. A project whose product is cryptographic proof of identity, built on the premise that you can trust math instead of trusting people, was brought down by the oldest trust problem there is. As CryptoSlate noted, despite the sophisticated cryptographic claims, Humanity still depended on ordinary custody controls. The vein maps and the zero-knowledge circuits were fine. The custody was not. The most advanced cryptography in the stack sat downstream of a plain text file on somebody’s hard drive, and the attacker never had to break the math because the math was never the lock.

One Laptop, Root Access

The breach began where a growing share of 2026 breaches begin: on an endpoint. Malware infected the computer of one member of the Humanity Foundation, gained root access, and began harvesting whatever it could find, including credentials, session tokens, and private keys stored on disk. This is the infostealer economy that has quietly reshaped crypto crime. An attacker does not need to beat your contracts if it can simply read your developer’s hard drive, and the tooling to do exactly that is now cheap, automated, and sold by subscription.

Halborn co-founder and chief technology officer Steven Walbroehl has warned repeatedly that this is where the fight now lives. Attackers, he told Cointelegraph, are using AI to build “highly personalized, context-aware attacks that bypass traditional security awareness training.” The weakest component in most crypto systems is not the virtual machine; it is the person with admin rights and an open browser. The October 2026 wave of wallet-targeting attacks that HOGE Wire catalogued earlier this month hammered the same theme from the consumer side: the device is the perimeter, and the device keeps losing.

What the malware found on that one machine was a jackpot. Not one key, which would have been bad enough, but seven.

The Backup Nobody Meant to Keep

Those seven keys were not supposed to be there. According to Kwok, the team had intended to disperse the keys properly, across separate people and devices, the way a multisig is meant to work. Instead, he said, “some of the keys were accidentally backed up to a compromised device during setup.” The setup in question was the protocol’s mainnet launch in June 2025. A backup made in the rush of going live was never cleaned up, and it sat on a working laptop for the better part of a year, waiting.

This is the quiet killer in key management. The policy was right. The architecture, on paper, was a multisig spread across owners. But a one-time convenience, a backup during a stressful launch, silently collapsed that architecture down to a single point of failure, and nobody went back to delete it. Security that depends on every engineer remembering to undo a shortcut is not security; it is a countdown with no visible clock. The failure was not a decision anyone made on 9 June 2026. It was a decision nobody unmade in June 2025.

Seven Keys, Two Safes, One Threshold

To see why seven keys were enough to take everything, look at how Humanity controlled its funds and its bridge. Two Gnosis Safe multisig wallets governed the critical machinery, one on Ethereum and one on BNB Chain, alongside an administrative hot wallet. On paper, moving anything of consequence required several independent owners to sign, the standard design that is supposed to make a single stolen key useless.

The laptop held three of the six owner keys for the Ethereum Safe and three of the five for the BNB Chain Safe, plus the hot-wallet key. Both Safes needed three signatures to act. Three signatures is exactly what the attacker now had on each chain, from one machine, without a single other signer lifting a finger. The multisig promise, that no one party can move funds alone, was void the instant three live owner keys shared a hard drive. Here is the attack laid out chain by chain.

The Drain

On Ethereum, the three Safe keys gave the attacker control of the bridge ProxyAdmin, the contract that can upgrade the bridge itself. With that, you do not need to find a bug in the bridge; you simply replace the bridge with your own code. The attacker transferred ownership to a wallet it controlled, deployed a malicious upgrade, and drained roughly 141 million H in a single transaction. There was no exploit in the usual sense, no clever reentrancy or rounding trick. There was an admin doing admin things, except the admin was a thief.

On BNB Chain, the attacker did not bother stealing existing tokens first. Using the second set of Safe keys, it installed what CoinDesk described as an unlimited-mint function that allowed the creation of tokens at will, and conjured hundreds of millions of fresh H directly into its own wallet. Across both chains, Halborn’s reconstruction puts the total at about 447 million H stolen or minted. When you own the contract that mints the money, you do not need to break in; you own the printing press.

Then came the part that turns a token exploit into a market event: the attacker sold. H was dumped across decentralized exchanges, and within a day the token had shed something like 80 percent to 90 percent of its value, trading near $0.17. The dollar figure that gets quoted, somewhere between about $31 million and $36 million depending on who is counting and when, is really the value the attacker extracted before the token it was denominated in collapsed under the weight of the sale.

Hack, or Heist From Within?

Not everyone believed the story at first. The on-chain investigator ZachXBT initially flagged the incident as “possibly staged,” pointing to suspicious market-making and over-the-counter selling ahead of a scheduled 25 June token unlock. In a market that has watched plenty of teams dress up an inside job as a hack, the skepticism was fair, and for a day or two it was the dominant read.

Then he traced the money and changed his mind. After following how the funds were laundered, roughly $23.7 million swapped into Ethereum while about $7.9 million stayed in H, ZachXBT concluded the pattern was inconsistent with a team quietly cashing out. “I thought that initially due to the active MM & recent OTC before unlocks,” he wrote, “however the evidence shared points to otherwise.” The suspicious trading and the breach, in his reading, were independent of one another and not related. The verdict: a genuine private-key compromise, not insider theft.

That distinction matters, and not only for Humanity’s reputation. It also shapes what can be recovered. No one has publicly tied this particular breach to North Korea; it carries the fingerprints of the commodity infostealer-malware economy rather than a nation-state operation, which is its own kind of warning. You do not need Lazarus to lose $36 million. You need one unpatched laptop and one forgotten backup.

A Multisig Is an Org Chart, Not a Spell

The most important sentence in Halborn’s post-mortem is also the least technical. “Web3 projects need security policies and programs that ensure that multisig private keys stay separated to avoid undermining their protection,” the firm concluded. Strip the jargon and it reads: a multisig only works if the keys actually live apart.

A Gnosis Safe set to three-of-five is a promise that moving money requires three different parties to agree. That promise, though, is social and operational, not cryptographic. The Safe contract cannot see whether its five owner keys sit on five hardware devices held by five people in five locations, or whether three of them are files in the same cloud backup. Cryptographically, the two look identical. Operationally, one is a multisig and the other is a single key wearing a costume. Humanity’s was the costume.

Real separation means hardware wallets or hardware security modules for each signer, signing on separate and ideally air-gapped machines, distribution across people and places, and an unbending rule that keys are never backed up to a general-purpose computer, not even once, not even on launch night. As HOGE Wire has argued in the context of consumer devices, the seed phrase is the leak: the signer’s environment, not the chain, is where keys are lost. Humanity’s failure was not that it lacked a multisig. It was that its multisig was a diagram, not a reality.

The Audit Reads the Code. It Cannot See the Laptop.

This is where the Humanity breach connects to Halborn’s day job, and to the limits of the entire audit industry. A smart-contract audit, Halborn’s or anyone else’s, examines contract logic: can this function be reentered, can this arithmetic overflow, can this access control be bypassed on-chain. It is a necessary discipline and it catches real, expensive bugs. What it does not do, and cannot do, is tell you whether three of your five Safe keys are backed up on an engineer’s laptop.

CertiK co-founder Ronghui Gu summarized the industry’s hardest truth to Forbes: “A protocol can pass a flawless code audit and still lose millions because of a compromised admin key.” Humanity is that sentence rendered in full. The contracts behaved exactly as written. The upgrade function worked; the mint function worked; the Safe required its three signatures and got them. Every line did its job. The job was simply handed to the wrong person. The gap between what an audit verifies and where protocols actually die is the whole story of 2026.

This is exactly why Halborn sells a Custody and Key Management Assessment and runs a full red-team practice alongside the audit. The code review answers whether the contract is correct. The red team answers the question Humanity needed answered: if I phish one of your engineers, how much can I take? In 2026, that second question is where the money is, and it is a question no amount of Solidity review will ever reach.

2026: The Year the Keys Became the Target

Humanity is not an outlier; it is a data point in an overwhelming trend. TRM Labs found that in the first half of 2026, infrastructure and operational compromises, meaning stolen keys, backend intrusions, and social engineering, accounted for only about 15 percent of incidents but roughly 76 percent of all value stolen. The common hack is still a contract bug. The expensive hack is almost never a contract bug. The money has moved to wherever the keys are, because that is where the money has always actually been.

Halborn’s own casebook says it louder. Kelp DAO lost about $292 million to a single-verifier bridge configuration and poisoned infrastructure, not a Solidity flaw. Drift lost roughly $285 million to a months-long social-engineering con that ended in pre-signed governance transactions. Bitget lost around $388 million in September when attackers compromised a third-party backend product and spoofed transaction data straight through the normal approval pipeline, a breach HOGE Wire reconstructed in detail at the time. In each case, the contracts were innocent. Humanity is the same story told in miniature, which is part of what makes it so useful: the mechanism is small enough to see end to end.

Then, two days before this was written, came the exclamation point. On 29 October, Binance lost about $6.36 billion, the largest crypto theft in history, in an attack chief executive Richard Teng attributed to an abused EIP-7702 delegation on a hot wallet, with TRM Labs tying the operation to North Korea’s TraderTraitor. EIP-7702 lets an ordinary account temporarily borrow a contract’s code; abuse it with the right signature and you redirect the wallet without ever touching the assets’ token contracts. Markets lurched, and the trading frenzy that followed became its own story, which HOGE Wire covered separately. Once again, the headline number came from the authorization layer, not the code layer.

The lone case in the recent record where the code genuinely was at fault is Liquid, whose roughly $320 million September loss traced to a flaw in the Elements rangeproof logic. It is the exception that proves the rule. September 2026 was, by multiple trackers, the worst month on record, north of $760 million gone, with Bitget and Liquid accounting for the bulk of it.

When There Is No One to Negotiate With

What happens after a hack like Humanity’s depends almost entirely on who the attacker is. When a whitehat or an opportunist drains a protocol, there is sometimes a deal to be made: return most of the funds, keep a bounty, walk away. When the attacker is a sanctioned nation-state, there is no deal, only tracing and freezing. Humanity’s case was neither a clean negotiation nor a clean attribution, which is the worst of both worlds for a victim.

In the hours after the drain, the team warned users not to interact with the protocol’s bridge or liquidity pools and said it was working with security firms and exchange partners to trace the funds. Kwok acknowledged the breach stemmed from compromised keys belonging to a member of the Humanity Foundation and said recovery would take time, because the team had to trace every transaction that followed the hack. By then the attacker had already converted the majority into Ethereum, the asset of choice precisely because it is hard to freeze, and the value was realized before most of the world knew there had been a hack at all.

The brutal arithmetic of recovery held. Realized losses from operational breaches like this one are rarely clawed back; the window to freeze funds closes in minutes, and an infostealer crew laundering into ETH is not going to answer a negotiation thread. There was no sanctioned issuer to call, no validator set to vote on a rollback, no whitehat with a conscience and a wallet address. For most of 2026’s victims in this category, the money is simply gone.