What happens when a hardware wallet that millions of people trust turns out to have been quietly modified before it even reached the customer? That question sat at the center of a heated public exchange this week between Ledger’s CEO and the former head of one of crypto’s most infamous exchanges. The numbers being discussed are not small. Independent researchers have pointed to more than $90 million in suspected losses tied to devices that looked completely genuine on the outside.
The Public Argument That Brought Old Wounds Back
On October 17 the conversation on X turned sharp. Mark Karpelès, once the public face of Mt. Gox, questioned whether Ledger had spoken clearly enough about a sophisticated form of physical tampering. Pascal Gauthier, Ledger’s chief executive, pushed back hard. He defended the company’s work and reminded everyone of Karpelès’s own history with a catastrophic exchange collapse more than a decade earlier.
The spark was a series of discoveries involving wallets sold through a Southeast Asian reseller. Investigators examining returned or reported devices found extra electronics that should never have been there. Those components sat near the screen and were capable of recording the 24-word recovery phrase during the initial setup process. Once captured, the phrase could be sent out over a cellular connection. No need for the attacker to touch the device again.
I’ve followed hardware wallet stories for years, and this one feels different. Most earlier attacks focused on fake devices or software tricks. This time the hardware itself was real, the security chip was authentic, and the device still passed the company’s own checks. That combination is what makes the whole episode so unsettling.
How the Dispute Started and Quickly Escalated
Karpelès initially claimed Ledger had stayed silent about the implant risk. He later corrected himself after realizing the company had published a security article months earlier. The piece, titled Best Practices To Securely Buy Your Ledger Signer, first appeared in March and received an update in September. It discussed counterfeit devices, pre-seeded recovery phrases, and the possibility of physical modifications that could capture information displayed on the screen.
Even after the correction, Karpelès argued that an educational article buried in the academy section was not enough. He suggested that a more prominent warning might have limited the scale of the losses. Gauthier rejected that framing. He stressed that Ledger invests heavily in protecting users and asked critics to stick to verified facts rather than speculation. The exchange grew personal when Gauthier referenced the Mt. Gox failure. Karpelès replied that surviving such a disaster made him more, not less, concerned about clear communication around customer funds.
Coming from Mt Gox CEO… Come on man. You are no white knight. Get real.
That line captured the tone. Old reputations and current responsibilities collided in public. Whether the earlier guide could realistically have prevented the thefts remains an open question. The article’s September revision also leaves uncertainty about exactly which images and technical details were present in the original March version.
What Investigators Actually Found Inside the Devices
The technical details are both clever and disturbing. The added components did not replace the Secure Element, the hardened chip that handles private keys and transaction signing. Instead they sat nearby and watched the screen during the one-time setup process when the recovery phrase is generated and displayed. That phrase is the master key. Anyone who obtains all 24 words can recreate the wallet and move the funds at any time.
Because the original security chip remained untouched, the device still passed Ledger’s Genuine Check. That process confirms the cryptographic identity of the Secure Element. It does not, and cannot, inspect every physical component or detect extra boards that leave the main chip alone. Ledger’s own documentation now states this limitation clearly.
The reported implants used cellular connectivity. That detail removes one of the classic constraints on physical attacks: the need for the thief to recover the device later. Once the phrase is transmitted, the attacker can act from anywhere. Funds can be drained while the legitimate owner still believes the wallet is safely in their possession.
In my view this is one of the more sophisticated supply-chain style attacks the industry has seen in recent years. It did not require compromising the manufacturer’s production line. It only required access to devices somewhere between the factory and the final customer, most likely at the reseller stage.
The Role of the Southeast Asian Reseller
Ledger began receiving complaints from customers who had bought devices through CryptoBilis. On October 9 the company asked the reseller to pause sales and shipments. Buyers who had purchased within the previous 90 days were told not to initialize any unused wallets. Those who had already set up devices were advised to move funds to a completely new recovery phrase generated on a trusted device.
The investigation later confirmed that at least one examined wallet contained unauthorized electronics capable of collecting the recovery information. Independent analysis, including work from Galaxy Research, placed suspected losses near $91 million. That figure is an estimate based on linked addresses and transaction patterns. It is not a final audited total of every device that may have been affected, nor is it proof that earlier public warnings would have changed the outcome.
Reports of a tampered wallet appearing in Germany prompted other retailers to review their own distribution channels. The incident quickly moved beyond a single reseller and into broader questions about how hardware wallets move through secondary markets and authorized partners.
Why Genuine Check Was Never Designed for This Threat
Many users assume that a successful Genuine Check means the device is completely trustworthy. That assumption is only partially correct. The check verifies that the Secure Element is authentic and was manufactured by Ledger. It does not open the case, scan for extra chips, or confirm that no one has added hardware after the device left the factory.
When the original security chip remains in place and continues to respond correctly, the software side of the verification succeeds. The physical implant can still perform its job. Ledger has now made this limitation explicit in its guidance. The company continues to recommend buying directly or through authorized channels, inspecting packaging for signs of tampering, and never using a recovery phrase that was already written down or displayed before the user generated it themselves.
Authorized resellers are also prohibited from returning previously sold devices to inventory for resale. That policy exists for exactly this kind of risk. Whether every partner follows it strictly is another matter, and one that the current investigation is still exploring.
The Timeline Raises Difficult Questions
The security article first appeared in March 2026. It was updated on September 18. The public investigation into CryptoBilis devices began in early October. The gap between the September update and the October revelations is only a few weeks, yet Karpelès and others have focused on the earlier March publication. They argue that if sophisticated implant designs were already being illustrated months earlier, the company should have raised the alarm more loudly once a specific distribution channel became suspect.
Ledger has not published a full revision history that would show exactly when particular images or technical descriptions were added. Without that record it is difficult to prove how much the company knew about the specific CryptoBilis modifications in March versus later in the year. Knowledge of a general attack method is not the same as confirmed awareness that compromised devices were already in a particular reseller’s stock.
This distinction matters. Publishing broad best-practice advice is responsible. Issuing a targeted warning about a live supply-chain problem is a different level of disclosure. The current public record does not settle which category the March article belongs to.
Practical Steps Users Should Take Right Now
Anyone who bought a hardware wallet through an unofficial or secondary channel in recent months should treat the device with caution. The safest approach is straightforward.
- Do not initialize any unused device purchased from a non-official source
- If the device was already set up, move funds immediately to a new wallet created on a trusted device
- Generate a completely fresh recovery phrase and never reuse the old one
- Prefer direct purchases from the manufacturer or clearly listed authorized partners
- Inspect packaging carefully and be suspicious of any signs of prior opening
These steps are not theoretical. They reflect the actual advice Ledger issued once the problem became clear. Moving funds to a new seed phrase is the only reliable way to break the link if the original phrase was captured.
I’ve spoken with several long-time users who keep multiple hardware wallets for exactly this kind of scenario. They treat the recovery phrase as the true point of vulnerability and assume that physical possession of the device can never be fully guaranteed once it leaves the factory.
Broader Lessons for Hardware Security
This episode forces a rethink of several assumptions that many in crypto have held for years. Hardware wallets are often described as the gold standard of self-custody. That reputation rests on the idea that the private keys never leave the protected chip and that physical attacks are difficult and expensive. Both statements remain true in the general case. They become less absolute when an attacker can insert extra electronics after manufacturing and before the customer ever sees the device.
Supply-chain integrity is harder to guarantee than cryptographic strength. A perfect Secure Element cannot protect against an implant that simply watches the screen during setup. The recovery phrase itself becomes the weak link. Once it is captured, the rest of the security architecture is irrelevant.
Manufacturers will likely respond with tighter controls on authorized distribution, more aggressive packaging seals, and clearer public education about the limits of device authentication. Some may explore physical anti-tamper measures that are harder to defeat without leaving visible evidence. Users, for their part, will need to treat the purchase channel with the same seriousness they currently apply to the seed phrase.
The Human Side of a Technical Failure
Behind the $90 million estimate are individual holders who believed they had done everything right. They bought what appeared to be genuine hardware, followed the setup instructions, and stored the recovery phrase carefully. The device passed the official check. Yet the funds still disappeared. That combination of careful behavior and sudden loss is particularly demoralizing.
The public argument between Gauthier and Karpelès has brought additional attention to the story. Whether that attention ultimately helps more users avoid the same fate remains to be seen. Public criticism can force clearer communication. It can also distract from the technical work still underway.
What is clear is that the industry cannot treat physical distribution as an afterthought. The strongest cryptographic chip is only as trustworthy as the path the device travels before it reaches the customer’s hands.
Looking Ahead: What Needs to Change
Several practical improvements seem inevitable. Manufacturers will expand the list of recommended purchase channels and may introduce more visible tamper-evident packaging. Software updates could add additional user education during the first setup flow, reminding people that a successful authentication check does not rule out physical modification. Independent researchers will continue examining devices from secondary markets, and that scrutiny is healthy.
For ordinary users the safest habits remain simple. Buy direct when possible. Inspect everything. Treat the recovery phrase as the single most important secret and generate it only on a device whose entire history is known. If any doubt arises, move the funds and start over. The cost of a new seed is trivial compared with the cost of discovering too late that the original one was compromised.
The clash between Ledger’s CEO and the former Mt. Gox executive will fade from the timeline. The underlying problem will not. Hardware wallets remain one of the better tools available for self-custody, but they are not magic. They operate in a physical world where determined attackers can still find gaps. Closing those gaps will require better distribution controls, clearer public communication, and a more realistic understanding of what device authentication can and cannot prove.
In the end the $90 million figure is less important than the lesson it carries. Trust in hardware is only as strong as the chain of custody that brings the device to the user. When that chain is broken, even the most carefully designed security chip can be outflanked by a small board and a cellular connection. The industry now has a vivid example of how that happens. The question is whether the response will match the seriousness of the threat.
Hardware wallet users who take self-custody seriously already know that the recovery phrase is the true point of failure. This latest episode simply underlines how early that failure can be engineered. The public argument between two high-profile figures has made the story more visible. The real work now is ensuring that fewer people ever have to learn the same lesson the hard way.
