A hardware wallet is supposed to be a signing device. One batch sold through CryptoBilis, according to Ledger, was a transmitter.
On October 13, Ledger confirmed that an unauthorized component inside affected devices read the seed phrase directly off the screen and pushed it out over a cellular network connection. Not a curve implementation flaw. Not a firmware exploit. A physical component with its own radio, inserted somewhere between the factory and the sealed box sitting on your desk.
The on-chain ledger of this reads: Bitquery counts 324 wallets drained, $102.1 million gone. Taylor Monahan is coordinating the report set with Ledger and receiving assistance from Trezor. The affected devices were purchased between June and early October.
Hold on that purchase window. It is the most load-bearing number in the entire story, and it is the one most coverage has already skated past.
The trust chain nobody audits
A hardware wallet's security argument has always rested on three assumptions stacked on top of each other. One: the secure element resists extraction. Two: the firmware signature check prevents unauthorized code. Three: the physical device you hold is the device that left the factory.
Ledger's cryptography holds up under the first two. The third one broke, and the industry has spent a decade treating it as an afterthought.
Trace the path: chip fab, assembly house, firmware flashing, packaging, distributor, reseller, courier, your hands. Every one of those junctions is a place where a component can be added or swapped. Ledger's disclosure points the confirmed losses at devices sold by CryptoBilis, a reseller channel rather than a vendor-direct storefront. That distinction matters enormously and almost nobody has priced it in. Code doesn't verify custody chains. A signature check validates bytes, not the journey the box took to reach you.
The contrast with Trezor is instructive but not flattering to either side. Trezor's open-source firmware means the software trust boundary can be independently examined. It says nothing about whether the plastic housing was opened before shipping. Monahan still runs both Ledger and Trezor — the changed variable is the purchase channel, not the brand.
That is the correct read, and it is a much narrower conclusion than the market took away.
What the screen actually leaks
Here is the part that deserves a diagram in every security bulletin.
During initialization, the seed phrase is rendered on the display. The 24 words appear in plaintext on a bus that connects the display controller to the main board. Anything with read access to that bus — an interposer on a flex connector, a substituted display module, a daughterboard spliced between the panel and the PCB — sees the seed in clear text, at the exact moment it exists in human-readable form.
I have a habit, drilled into me during my 2021 audit of NFT marketplace approval flows, of tracing where plaintext crosses a privilege boundary. In that case the finding was that marketplace approval logic handed broad minting rights to callers that should never have held them. The lesson generalizes: the cryptographic layer is rarely where the body is buried. It is almost always the interface that hands plaintext to something with more privilege than it needs.
A hardware wallet that displays a seed phrase on a screen has, by design, a plaintext exposure window. The secure element protects the key at rest. It does not protect the words during the thirty seconds a user writes them down.
Now add the radio.
A conventional skimmer requires retrieval — the attacker has to get the device back, or get the user to plug it into a network. A cellular module changes the class of attack entirely. The device phones home. Exfiltration becomes near-real-time, which means the attacker knows the moment the wallet is funded, which means the drain can be scheduled rather than stumbled into. Code doesn't need your password if the hardware you bought is already talking to someone else.
The second wave is larger than the first
Here is the finding that has been underweighted in nearly every writeup I have read this week.
Monahan's most recent batch of reports is not implant victims. It is people who panicked. Malicious Google ads. Fake apps. Phishing sites. Threatening emails. Phone calls claiming to be Ledger support.
SlowMist warned on October 16 about phishing pages impersonating a Ledger firmware update. Zscaler ThreatLabz documented fraudulent Google ad campaigns of exactly this shape in September 2024 — a full year before this incident. The infrastructure was already built. This event simply supplied the pretext.
Ledger Support posted on October 14 that Ledger will never call you, never DM you, never ask for your 24-word phrase. Correct crisis communication. It did not stop the copycats, because copycats do not need the brand's permission to use the brand's name.
And the attribution problem sits on top of all of it. Bitquery's 324 wallets and $102.1 million is a real number, but it is an aggregate across attack paths, not a clean measurement of the implant. Ledger has not published a confirmed loss total. Monahan has not published a case count. Two missing numbers, both material, both currently filled in by speculation.
Since the Terra post-mortem I published in 2022, I run every incident through the failure question before the success question: what does this look like if it goes wrong? Applied here, the answer is uncomfortable. The failure mode is not a compromised Ledger. The failure mode is a user who concludes the danger has passed because the official statement scoped confirmed losses to one reseller.
The contrarian read: the exploit doesn't scale, the panic does
An implant requires physical access to a distribution channel, a component, a SIM, an operator, and a payment path. It is expensive, geographically constrained, and finite.
A phishing page requires a two-dollar domain and a sponsored ad slot.
One of these has a ceiling. The other does not. Monahan telling people not to panic is not a sentiment — it is a security control, and it is the highest-leverage control in this entire incident.
The second thing nobody wants to say out loud: multisig has been technically available for years and is still being framed as a hedge rather than a default. Austin Griffith's point is straightforward — a hardware wallet can be one of N signers, and if it is lost or leaked, you replace one signer and nothing else moves. Monahan's own practice, multiple devices, multiple locations, paper backups distributed, is a manual version of the same idea.
It has not become standard because single-device self-custody is a better product story. "Not your keys, not your coins" fits on a box. "Your keys, split across three signers in three locations, with a paper backup you must actually maintain" does not.
The third blind spot is structural. If the weak link is a distributor, the fix is channel attestation: cryptographic proof of manufacture, seal verification at unboxing, a verifiable chain-of-custody record. That product does not exist as an industry standard. Hardware vendors compete on secure elements and open firmware because both are visible on a spec sheet. Nobody competes on supply-chain attestation, because it is expensive and invisible until the day it is everything.
I unbox hardware wallets on camera and photograph the seal. I have done that since 2021, and until this week I thought of it as theatre. It is not theatre. It is the only control I personally hold over a link in the chain that no vendor has ever offered to prove to me.
What to watch
Three signals over the next quarter. Whether Ledger ships a device-verification or channel-attestation tool, and whether it is verifiable rather than merely announced. Whether multisig onboarding volumes move, because that is the only durable behavioral change this incident can produce. Whether Google tightens sponsored-ad verification for crypto brands, since Zscaler's September 2024 writeup suggests the ad channel has been open for a long time.
And the question that should be on every hardware vendor's roadmap meeting this week: if a component you never inspected can read your seed phrase off your own screen and mail it somewhere, what exactly did you verify when you verified the firmware signature?
