The implant in Ledger cold wallets: how the reseller attack works, whether a scale helps, and what it means for Trezor
Data as of 14.10.2026.
Short version. There is no public evidence that anyone has planted an implant in a Trezor device (Trezor, 13.10.2026). The Trezor Safe 3 weakness, known since March 2025, involves the same threat as the Ledger case: tampering with a device before it reaches the buyer. So Safe 3 risk depends on where the device came from, not on the brand. A device bought directly from Trezor or an authorised reseller, with an intact seal and a recovery phrase generated on the device itself, carries no new risk because of the Ledger incident. If the origin can't be verified, it's the same class of risk the CryptoBilis buyers face, and the fix is the same: a new device from an official channel, a new phrase, and moving the funds. Weighing the device on a kitchen scale doesn't answer the question.
What it is
A 12.10.2026 post from the DEFI Scam Check channel suggests weighing your hardware wallet and comparing it with the spec-sheet weight. The trigger is a physical implant that Ledger confirmed on 10.10.2026 in devices sold by the reseller CryptoBilis in Southeast Asia. Loss estimates range from $86M to $103.4M. The post's X link (0xcramjam) can't be read automatically (HTTP 402, 14.10), so its content isn't used here. "Weigh your wallet" is treated below as a claim to be tested.
What's new and how it works
The attack (Ledger Donjon, disclosed 10–13.10.2026). A separate microcontroller sits inside the case and reads whatever is on the screen. During setup the device displays the 24 words of the recovery phrase. The implant recognises those screens, decodes the words and sends them out over LTE through its own preconfigured eSIM. It doesn't touch the Secure Element, Ledger OS, transactions or signatures (The Block, 13.10; Gizmodo; Forklog, 13.10).
Why this is a new class of attack rather than another hack: 1. "Cold" doesn't help. The implant has its own cellular link, so air-gapping, QR signing or switching off Bluetooth make no difference. The phrase bypasses the computer and the phone entirely. 2. The standard authenticity check can't see it. That check confirms the Secure Element and firmware are genuine, and the implant modifies neither. The device passes every vendor check and still leaks the phrase. 3. The attack goes through the reseller, not the manufacturer. Ledger says its systems were not breached and that confirmed losses involve devices sold by CryptoBilis (FinanceFeeds, 12.10; 24/7 Wall St, 12.10). Changpeng Zhao claims the attackers bought the reseller outright. Neither Ledger nor CryptoBilis has confirmed this (same sources).
Scope. Models named in the sources: Nano X, Flex, Nano S Plus, Stax (Gizmodo). A separate case in Germany involved one device and no lost funds. After it, MediaMarkt/Saturn pulled Ledger from sale in Europe (13.10). A buyer's claim of two more devices from different MediaMarkt stores is unverified (The Block, 13.10).
Losses: the sources disagree, and the disagreement is itself a finding. Ledger: "figures circulating online come from third parties and have not been fully verified by Ledger" (The Block, 13.10). The spread of $86–103.4M (×1.2) and 315–530 addresses (×1.7) shows that trackers attribute addresses to this attack differently. Treat the loss as a range, not a single number.
Does weighing work? The post's Trezor weights (Safe 3 — 14 g, Safe 5 — 23 g, Safe 7 — 45 g) match the official specs (trezor.io). The Nano X at 34 g also matches (shop.ledger.com). Nano S Plus 21 g, Flex 58 g and Keystone 3 Pro 103 g are unverified. The test is weak for three reasons: - the implant's mass (microcontroller, LTE module, eSIM, wiring) hasn't been published anywhere: not in Ledger's X update, The Block, Forklog or Gizmodo. Nobody knows how many grams to look for; - spec weights are rounded to the gram, and kitchen scales typically have an error of about ±1 g (typical for the device class, no specific source). On a 14 g device that's ±7%; - the test only works one way. An extra 2–3 g on a jewellery scale (0.01 g resolution) is a reason to stop, but a "normal" weight proves nothing. An attacker could offset the mass, e.g. with a smaller battery in models that have one (inferred from the mechanism, not verified).
Neither Ledger nor Trezor recommends weighing as a check. It doesn't appear in the sources above or on Trezor's "Is my device safe to use?" page (trezor.io). A scale is a cheap extra filter, not a verification.
What it means
- The main risk is the channel the device came through, not the brand. The attack didn't break Ledger's cryptography. It swapped the device before it reached the buyer. The same route works against any brand whenever the chain includes a reseller that can be bought or bribed. The market has learned this lesson with software tooling before: on 20.09.2022 Wintermute lost $160M to a weak seed generator in Profanity, which made the tooling supply chain its own attack surface. The lesson now extends to retail hardware.
- One phrase on one device is a single point of failure. The implant shows it in practice: anyone who generated a phrase on an infected device lost everything behind it. Splitting custody (several phrases, multisig, a passphrase) limits the damage from a single compromise, because one leak is no longer enough.
- The market's response is to shrink distribution. Trezor removed CryptoBilis from its reseller list (11–12.10) and on 13.10 announced it is cutting its reseller network, with some authorisations ending on 01.01.2027. Trezor: "there is no evidence that Trezor devices have been compromised" (Crowdfund Insider, 10.2026; Forklog, 13.10). OneKey is replacing devices bought from CryptoBilis between 01.03 and 13.10.2026 free of charge (The Block, 13.10). As a result, buying direct from the manufacturer is becoming the norm, and "cheaper from a local seller" is now a risk in its own right.
- Phishing follows the incident. On 14.10, search results for a "Ledger verification guide" include daftpage.com pages with URLs that mimic ledger.com (not opened). Any "device check" that asks for the phrase on a website is theft.
- Background: theft from individuals is rising. Per Chainalysis, personal wallets accounted for 23.35% of stolen funds in H1 2025, and physical-coercion ("wrench") attacks were on track to double the record (The Block / Chainalysis, 07.2025). In the full-year report (12.2025) the share is about 20%, across 158k incidents. Attackers increasingly target the person and their device, not the protocol.
- For Ledger as a company, this is a reputational hit in the year it is preparing an IPO targeting a valuation above $4bn (MarketScreener).
Base rate. 315–530 addresses against 7.2–8M Ledger devices sold is ≈0.004–0.007%. But an address isn't a device, and the rate within the CryptoBilis channel alone is unknown, so for buyers from that reseller the risk is orders of magnitude above the average.
Market stage and macro
Market stage: the hardware wallet market is scaling (Ledger has sold 7.2–8M+ devices, and the sources disagree: CoinLaw, yespress), while physical supply-chain implants are at early detection: this is the first mass case confirmed by a manufacturer (Ledger, 10.10.2026), there is no base rate yet, so one case can't be extrapolated to other brands; confidence 0.7. Macro: the macro regime is supportive but fragile (reading of 19.07.2026, close to going stale). Macro affects this topic only through price, since attacks on crypto holders correlate with the bitcoin price (Chainalysis, 07.2025). Timing: on time — the incident is still unfolding, resellers are being purged right now, and checking the purchase channel takes 10 minutes.
Is the Trezor Safe 3 at risk?
What is known: - Secure Element OPTIGA Trust M (V3), CC EAL6+ certified; Cortex M4 microcontroller, 180 MHz; weight 14 g (trezor.io); - March 2025: Ledger Donjon showed that the Safe 3's firmware integrity check can be bypassed by desoldering the microcontroller and applying a precisely timed voltage glitch. The attack needs physical access, equipment and skill. According to Trezor, the researchers did not extract keys or the PIN (The Block, 03.2025; Cointelegraph, 03.2025). Donjon itself framed it as a bypass of the protection against supply-chain attacks specifically; - a firmware update doesn't fix it (Trezor: "unfortunately, no", Cointelegraph).
What follows. The Safe 3 weakness fits the same scenario that hit CryptoBilis buyers: someone gets the device before the buyer does. Once the holder has set it up, the flaw doesn't expose the keys (per Trezor). So the answer depends on where the device came from: - bought directly from Trezor, its official Amazon storefront or an authorised reseller; seal intact; no firmware on first connection; phrase generated by the device — the Ledger incident adds no risk. The Donjon scenario remains theoretical; - bought second-hand, on a marketplace, from an unauthorised seller or from CryptoBilis, or the circumstances are unknown — the risk is real, and higher for the Safe 3 than for the Safe 5 and Safe 7. The fix is what Ledger advises its own buyers: a new device from an official channel, a new phrase, and moving the funds (FinanceFeeds, 12.10); - in every case, a passphrase (the so-called 25th word) raises the bar. Against the Kraken attack on the Model One and Model T, a strong passphrase gives full protection (Kraken, 31.01.2020). Against a Ledger-style implant it helps only partly: the 24 words from the setup screens don't open the hidden wallet without the passphrase (inferred from the mechanism). But an implant that reads the screen could in theory read the passphrase too, if it's entered on the device itself.
All Trezor models: physical attacks
| Model | Secure chip | Known physical attack | What protects | Weight |
|---|---|---|---|---|
| Model One | none | Kraken, 31.01.2020: encrypted seed extracted via glitching with 15 min of physical access; hardware flaw, not patchable | a strong passphrase fully closes the attack | unknown |
| Model T | none | the same Kraken attack | strong passphrase | unknown |
| Safe 3 | OPTIGA Trust M V3, EAL6+ | Donjon, 03.2025: firmware integrity check bypassed via the microcontroller; needs physical access and disassembly; keys not extracted; not fixable by firmware | buy only through official channels, seal, passphrase | 14 g |
| Safe 5 | OPTIGA Trust M V3, EAL6+ | not affected per The Block (STM32U5); Cointelegraph mentions Safe 3 and 5, so the sources conflict | same | 23 g |
| Safe 7 | OPTIGA Trust M V3 plus a second chip, TROPIC01 | no public attacks found (search 14.10) | same | 45 g |
Spec sources: trezor.io; attacks: Kraken, The Block, Cointelegraph.
Checklist, from strongest protection to weakest: 1. Purchase channel. Only the manufacturer's shop, its official storefront, or a seller on the manufacturer's reseller list, checked on the day of purchase. A device of any brand from CryptoBilis means a new device, a new phrase and moving the funds. 2. The phrase comes only from the device. A phrase in the box, on a card or from the seller, or a device asking for a "ready-made" phrase, means theft (24/7 Wall St, 12.10). 3. Seal and blank firmware. Trezor ships devices without firmware. If firmware is already installed on first connection, someone has connected the device before (trezor.io). 4. Use the Trezor Suite authenticity check, but know where it stops. It catches a swapped chip or firmware. It can't see an implant that only reads the screen: in the Ledger case, a genuine Secure Element and firmware didn't stop the theft. 5. Passphrase. It closes the Kraken attack on older models, and a leak of the 24 words alone no longer opens the wallet. 6. Splitting. Multisig, or several phrases on devices from different batches and channels: one infected device shouldn't be able to give everything away. 7. The scale comes last. 2 g or more over spec on a 0.01 g scale is a reason to stop. A "normal" weight proves nothing. 8. After a loss, don't open the device yourself. Ledger advises going to law enforcement first (The Block, 13.10). 9. No "verification websites". Clones of vendor pages appear after an incident. The phrase is entered nowhere except on the device itself.
What we don't know
Unknown or unpublished: - the implant's mass and physical size: not published in Ledger Donjon's X update, The Block, Forklog, Gizmodo or FinanceFeeds; - the real loss and number of victims: Ledger hasn't confirmed any estimate, and trackers give $86–103.4M and 315–530 addresses; - how the attackers gained control of the reseller: Zhao's version is unconfirmed; - whether implants exist outside the CryptoBilis channel: one device in Germany (MediaMarkt) is confirmed, other claims are unverified; - whether a jewellery scale can detect the implant: no infected sample is publicly available, so the answer has to wait for a detailed Donjon teardown.
Data exists but is unverified: - Ledger's official device-verification guide (the link in the coverage didn't open automatically); - whether the Safe 3 issue is fixed in new batches: Trezor says it is "resolved", but it can't be fixed by a firmware update (Cointelegraph, 03.2025), so this needs a Trezor bulletin on a hardware revision; - whether the Donjon finding also covered the Safe 5: The Block says no, since it uses a glitch-resistant STM32U5 microcontroller (The Block, 03.2025), while Cointelegraph mentions "Safe 3 and 5". The original Donjon post would settle it; - the post's weights for the Nano S Plus, Flex and Keystone 3 Pro haven't been checked against the vendors' spec pages; - the 0xcramjam post on X couldn't be read.
Single source: a mention of a 2026 Coldcard firmware vulnerability with losses above $100M (Crowdfund Insider) hasn't been confirmed by a second source and isn't used in the conclusions.
What would change the conclusion
(1) A confirmed implant in any Trezor device; (2) implants in direct-from-manufacturer sales channels. Either would make the attack systemic and prove "the risk is the channel, not the brand" wrong. (3) A Donjon publication giving the implant's mass, which would turn the scale into a real test.
| Date | Subject | What to check | Where | Expectation |
|---|---|---|---|---|
| 14.12.2026 | Implants in Trezor devices | Has Trezor or an independent researcher confirmed a physical implant in at least one Trezor device | trezor.io/blog, Trezor on X, The Block, Forklog | No confirmed case |
| 14.12.2026 | Distribution channels of Ledger implants | List of retail channels where Ledger or Donjon confirmed an implant | Ledger and Donjon on X, ledger.com, The Block | At most two retail channels: CryptoBilis and one more |
| 14.12.2026 | Weighing as a test | Does device weight appear in official Ledger or Trezor verification guidance or in the Donjon teardown | ledger.com support, trezor.io support, donjon.ledger.com | No manufacturer adds weighing to its official verification guide |
Research for information only. Not investment advice.