Ledger, Mt. Gox, and a $90M Dispute: What Trusting a Hardware Wallet Actually Means

Two of the more recognisable names in crypto are arguing in public, and the number attached to the argument is roughly ninety million dollars.

On one side is the chief executive of Ledger, the French company that makes one of the most widely used hardware wallets in the world. On the other is Mark Karpeles, who ran the Mt. Gox bitcoin exchange until it collapsed in 2014 with around 850,000 customer bitcoin missing. As reported by beincrypto.com, the two have clashed publicly over a wallet-related security incident, with language in the coverage referencing allegations of a hardware "spy implant" — a claim that, at the time of writing, has not been independently confirmed.

You can read the original news item here: Ledger CEO and Former Mt. Gox Head Dispute Over $90M Wallet Security Incident.

Before going further, the honest part: the specifics are murky. It is not clear from available reporting what exactly happened, who lost what, or whether any device was actually tampered with. Both participants have reasons to want their version to be the one people remember. Treat every claim in this story as unverified until someone independent verifies it.

What the dispute is genuinely useful for is this: it forces a question most people never ask. When you buy a hardware wallet and move your coins onto it, who exactly are you trusting, and for what?

That question has a real answer, and it will still be relevant long after this particular argument is forgotten.

What a hardware wallet actually protects

Start with the thing being protected. A wallet in crypto does not hold coins. The coins — the balances — live on the blockchain. What the wallet holds is the private key: a secret number that proves you are allowed to move those balances.

Anyone with the private key can spend the funds. Nobody without it can. There is no password reset.

A hardware wallet is a small dedicated device, usually the size of a USB stick, built to do one job: generate a private key, keep it inside the device forever, and sign transactions without ever letting the key out. You plug it into a laptop or phone, the computer builds a transaction, the device signs it, and the signature comes back. The key itself never touches the internet-connected machine.

That design exists because of a specific threat: malware on your laptop. If your private key is sitting in a file or a browser extension on a machine that also downloads things, opens attachments, and runs dozens of programs, a single piece of malware can read it and drain you in seconds. The hardware wallet removes the key from that environment entirely.

It solves that problem well. It does not solve every problem, and that is where disputes like this one become interesting.

The trust stack nobody reads

When you use a hardware wallet, you are not trusting one thing. You are trusting a stack of things, each of which could fail independently.

The chip. Most reputable hardware wallets use a secure element — a tamper-resistant chip of the kind used in bank cards and passports. You are trusting the chipmaker, and you are trusting that the chip's firmware does what it claims.

The firmware. This is the software running on the device. You are trusting that it generates keys randomly and unpredictably, that it signs only what you approved on the device screen, and that it has no hidden path for extracting the key.

The companion app. The desktop or mobile software that talks to the device. You are trusting it to show you the correct destination address and amount. This is why hardware wallets have their own screens: you are supposed to verify the address on the device, not in the app, precisely because the app runs on a machine that might be compromised.

The supply chain. The factory, the warehouse, the courier, the reseller. You are trusting that the sealed box you opened contains the device the manufacturer shipped and nothing else.

The company itself. Its data handling, its business decisions, its response when something goes wrong.

Ledger has been tested on several of these layers before. In 2020, a breach of its e-commerce database exposed customer names, addresses and phone numbers — not keys, but enough to fuel years of targeted phishing and, in some reported cases, physical threats. In 2023, the announcement of an optional key-backup service called Ledger Recover prompted a loud argument about whether a device marketed as never letting the key leave could meaningfully offer a feature that exports a key at all. Neither of those events was a break of the cryptography. Both were arguments about where the trust boundary sits.

That is the pattern worth internalising. Hardware wallet failures are rarely "the maths broke." They are almost always about a layer the user did not know they were trusting.

Supply chain attacks, explained plainly

The "implant" framing in this story points at the layer most people never think about: the physical journey from factory to your hands.

A supply chain attack means the attacker gets to the product before you do. In crypto, this has taken several documented forms over the years:

• Pre-seeded devices. A reseller opens a device, sets it up, writes the recovery phrase on the enclosed card, reseals the box, and sells it. The buyer follows the card, loads funds, and the seller — who also has the phrase — empties the wallet later. This is the most common version and it needs no technical skill at all.
• Scratch-card scams. A fake "your recovery phrase" card with a scratch-off panel, designed to look official.
• Modified hardware. Replacing or adding a component so the device leaks keys or signs transactions the user did not approve. This is far harder, far more expensive, and far rarer than the first two.
• Fake firmware prompts. Software that imitates a genuine update and asks you to type your recovery phrase into a computer. No legitimate device will ever ask for this.

Allegations of deliberate hardware tampering are serious and also genuinely difficult to prove. Demonstrating that a chip was modified in transit requires physical forensics on the specific unit, and even then, establishing where in the chain it happened is a separate problem. That difficulty is exactly why claims like this tend to produce long public arguments rather than quick resolutions.

What this translates into in practice

None of the following is advice about what to own. It is simply what the threat model above implies about handling a device.

• Buy from the manufacturer or an official channel, not a marketplace listing or a used-goods site.
• A new device should generate your recovery phrase in front of you, on its own screen. If a phrase comes pre-written in the box, the device is not yours.
• Write the phrase down yourself. Never type it into a computer, a phone, a photo, or a support chat.
• Run the manufacturer's genuineness check when you first connect the device.
• Verify every receiving and sending address on the device screen, not in the app window.
• Treat the amount you keep on any single device as a decision, not a default.

If you want to walk through the setup carefully rather than guessing, our Wallet Setup Walkthrough covers the first-device process step by step, including the checks that catch tampering before funds are at risk.

The Mt. Gox half of the story

The presence of Karpeles in this dispute is a reminder of the other failure mode entirely.

Mt. Gox was a Tokyo exchange that at its peak handled the majority of global bitcoin trading. Its customers did not hold private keys. The exchange did. When the coins went missing in 2014, users discovered their balance was never a balance — it was an IOU from a company that could not pay. Claimants have been waiting over a decade for partial distributions.

So the two sides of this argument represent the two categories of crypto risk, not one:

Self-custody does not eliminate risk. It changes who carries it, and it changes what kind of mistake is fatal. Someone moving funds off an exchange because of Mt. Gox and onto a device without understanding supply chain checks has traded one unexamined trust for another.

Why this keeps getting more important

The trust question is widening, not narrowing. Software now increasingly holds signing authority on a user's behalf — automated trading programs, smart contract wallets, and more recently AI agents given spending permissions. Each of these adds another layer to the stack above. Our lesson on agent wallets and autonomous onchain actors looks at what it means to delegate signing power to software, and what guardrails exist when you do.

The underlying discipline is the same whether the signer is a metal device in a drawer or a program running in a data centre: know what holds the key, know what could make it reveal that key, and know what happens to you if it does.

What to watch in this specific story

If you follow this dispute further, the useful signals are narrow ones:

• Has any independent party examined a physical device and published findings?
• Is there an onchain trail for the funds in question, and does it match either account?
• Are the claims about a manufactured product, or about one specific incident involving one specific person?

Until those are answered, it is a disagreement between two parties with strong incentives, not a finding. The concept underneath — that a wallet is only as trustworthy as the least-examined layer beneath it — is true regardless of how the argument ends.

This article is educational and is not financial advice. CryptoBipto does not hold funds, execute trades, or recommend products.