Hardware Wallet Reviews 2026: Grade the Channel, Not the Chip
Trezor confirmed the CryptoBilis implants and shipped a firmware fix on 11 November, but it cannot recover a seed already stolen. Here is how to review a hardware wallet in 2026.
On 11 November 2026, Trezor pushed firmware 2.13.1 to its Safe 7, Safe 5, Safe 3 and Model T. The release carried two sentences that, read together, rewrite the job of reviewing a hardware wallet. The first was meant to reassure: the update, Trezor said, “includes a protection that stops the implants we’ve found from capturing anything on devices that run it, making the current batch of implants useless.” The second was not meant to reassure anyone: “the update cannot protect wallet recovery phrases that may have already been captured before it was installed.”
The implants were not theoretical. Across October and November, both of the two best-known hardware wallet makers confirmed that devices sold through one Southeast Asian reseller, CryptoBilis, had been fitted with a hidden circuit board wedged between the main board and the screen. The board reads the recovery phrase as the device displays it, then texts it out over a cellular connection to the thief. Ledger confirmed its version in October; Trezor’s own laboratory confirmed three tampered Safe 5 units on 11 November. Neither implant touched the secure element. Neither altered a signature. They did not need to.
That is the uncomfortable lesson of 2026, and it lands in a brutal market. Bitcoin traded near $61,300 on the morning of 20 November, down about 3.8 percent on the day and roughly 29 percent below where it sat a year earlier, according to Fortune. Falling markets push people toward self-custody, whether to cut counterparty risk after a year of exchange stress or simply to stop watching a balance drop on someone else’s app. The reflex is healthy. The old way of choosing a device is not. For most of the last decade a hardware wallet review was a datasheet contest: whose secure element carried the higher certification grade, whose firmware was open source, whose screen was sharper. In late 2026 the secure element is the one part of the system nobody managed to break. The weak links were the supply chain, the vendor’s customer data, and the plain fact that most of these devices show your 24 words on a screen the day you set them up. A review now has to grade the channel, not just the chip.
The Review Question Moved From the Chip to the Channel
Line up every hardware wallet disaster of 2026 and a pattern jumps out: not one of them was a broken secure element. A five-year-old build flag turned Coldcard’s random number generator weak, draining wallets whose seeds were generated with as little as 40 bits of real entropy. A third-party fulfillment vendor leaked tens of thousands of Trezor customers’ names and shipping addresses, fuel for targeted phishing. A reseller shipped Ledgers and then Trezors with spy chips soldered behind the display. And across the whole year, thieves skipped the electronics entirely and showed up at people’s homes with threats. The chips held. Everything around them did not.
So a modern review cannot stop at the spec sheet. It has to grade five layers in order: the secure element and who is allowed to audit it; the firmware, its build reproducibility and its entropy; what you are actually approving when you sign (blind payload or human-readable transaction); the physical and data path the device travels to reach you, plus how the vendor behaves when that path is breached; and finally your own threat model, because a device graded for a casual saver is wrong for someone with a public seven-figure balance. The chip is layer one of five, and it is the layer that already works. The three in the middle are where 2026 did its damage, and the last one is the one no manufacturer can sell you.
What Actually Happened: The CryptoBilis Timeline
The story that reset the category began quietly. On-chain investigators later found that the attacker set up control addresses in late September and ran a fortnight of small test transfers before the main event. Then, on 9 October, large unexplained drains started surfacing, and every confirmed victim had one thing in common: they had bought their device from CryptoBilis, an authorized Ledger reseller serving Indonesia, Malaysia and the Philippines. Ledger confirmed on 10 October that at least one impacted device “contained an unauthorized hardware implant.” A teardown found a tiny microcontroller, a 2x2mm eSIM and a shielded LTE module hidden behind the Nano X display, wired to read the 24 recovery words as they appeared on screen during setup and transmit them over the mobile network. The personal computer was never involved; the attestation “genuine check” passed because the secure element itself was untouched and the implant simply sat beside it.
The money was real and the tally kept moving. On-chain analytics firm Bitquery traced more than $100 million drained from over 300 wallets across six chains, while Arkham put it near $95 million and mainstream coverage settled around $86 million to $93 million as figures were reconciled; Tether froze roughly $10 million of the stolen stablecoins. Then the story jumped brands. In late October an Indonesian customer reported a Trezor Safe 5, bought through a marketplace listing, that carried a similar implant, and Trezor issued a public warning on 30 October. Twelve days later its lab confirmed the worst. The secure element was never the target. The box it shipped in was.
The Firmware Fix, and the Fine Print
Trezor’s 11 November response is a case study in what a good vendor reaction looks like, and in the limits of software. The company examined the original reported device in its laboratory and identified two more compromised wallets from photographs their owners sent in; all three were Safe 5 units, all bought from CryptoBilis between 20 August and 23 September. Firmware 2.13.1 ships a protection that, in Trezor’s words, makes the current batch of implants “useless” on any device that installs it. The implant sat between the main board and the display and changed nothing on the wallet itself, so the fix works by starving it of the screen data it was built to read.
Here is the sentence a buyer has to internalize: the patch cannot recover a recovery phrase that was already captured. If a tampered device showed you your 24 words, and an LTE module beside the screen read them, those words left the building weeks ago; a later update does not call them back. That is why Trezor’s advice is not “install the update and relax” but “move your funds to a new wallet with a new recovery phrase.” Firmware fixes the device. It cannot fix a secret that has already leaked, which is exactly the property that makes a hardware wallet valuable and also unforgiving.
The structural response matters more than the patch. Trezor said it cut its global reseller network by roughly 80 percent and will apply stricter sourcing, storage and access rules to the partners that remain, and it offered free replacement devices to anyone who bought a Trezor from CryptoBilis since March 2026, including customers whose units were never confirmed as tampered. In a letter to partners the company put the lesson plainly: a hardware wallet is “only as safe as the path it travels to reach its owner.” That line, not the firmware changelog, is the review criterion for 2026.
Why the Secure Element Was Never the Weak Link
It is worth dwelling on how hard these chips actually are, because the reflex after a hack is to distrust the hardware, and here the hardware is the strong part. A secure element is a tamper-resistant chip certified against Common Criteria grades such as EAL5+ and EAL6+, designed to resist physical probing, side-channel leakage and fault injection. Ledger builds on ST33-series elements running its closed-source BOLOS operating system; Trezor’s flagship Safe 7 pairs an Infineon element with TROPIC01, billed as the first openly auditable secure element, while the cheaper Safe 3 and Safe 5 use Infineon’s Optiga Trust M; Tangem seals a Samsung element into a card and generates keys on it. These are not the parts that failed.
The proof is that even the researchers who did bend a chip needed a laboratory to do it. In 2026 Ledger’s own Donjon security lab ran laser fault-injection attacks against both rivals. Against a Tangem card it fired a nanosecond laser pulse that faulted a single conditional check in the card’s set-PIN routine, resetting the access password without the original. Against Trezor’s TROPIC01 it defeated one of three independent layers. Both attacks required physical possession of the device, a rig costing around $250,000, and hours of work per unit, and both produced exactly zero real-world losses. Compare that to the CryptoBilis implant, which needed a cheap board and a shipping label. When your most sophisticated chip break costs a quarter of a million dollars and steals nothing, while a reseller with a soldering iron steals nine figures, you are reviewing the wrong layer if you only read datasheets.
Both targeted vendors made the same point. Tangem co-founder Andrey Kurennykh argued in a public rebuttal that “LFI attacks are not scalable, and for everyday users, the practical risk is virtually non-existent,” noting there have been no known real-world losses from laser fault injection on any hardware wallet. Trezor chief executive Matej Žák said of the TROPIC01 finding that “a vulnerability in TROPIC01 does not put user funds at risk,” because the Safe 7 was built with multiple independent layers. Both are right, and both answers point away from the chip and toward the system around it.
The Five Layers a 2026 Review Has to Grade
If you want a checklist rather than a vibe, here is the one this piece uses for every device below. Each layer is a question, and the answers are not all available on a spec sheet.
- The chip. Does it use a certified secure element at EAL5+ or higher, and who is allowed to audit it? A sealed, certified element is table stakes; an openly auditable one is a bonus, not a requirement, as 2026 proved by failing everywhere except here.
- Firmware and entropy. Is the firmware open source with reproducible builds, so independent parties can confirm the code on the device matches the published source? Where does the seed’s randomness come from, and has that path been audited? Coldcard’s $116 million disaster was an entropy bug, not a chip flaw.
- Signing. When you approve a transaction, do you see a human-readable summary, or an opaque payload you are trusting blindly? Blind signing has been tied to some of the largest thefts of the era.
- The channel. Where did the device physically come from, and how does the vendor behave when that channel is breached? Direct from the maker beats a random marketplace listing, and a vendor that cuts resellers, ships a fix and offers replacements is passing the test the CryptoBilis buyers failed.
- You. What is your actual threat model? A casual saver and a publicly doxxed holder need different devices and, more importantly, different habits. No feature on any box answers a wrench.
Hold those five in mind through the device notes. The spec sheet speaks to the first three. The last two are where real money was lost this year.
The Field at a Glance
List prices below are the manufacturers’ own, taken from the Ledger and Trezor stores and Tangem’s shop; street prices drift with the Bitcoin demand cycle, and the whole category still sells more units in bull years. One number worth keeping in view: industry trackers estimate the hardware wallet market at well under a billion dollars in annual sales today, a rounding error against the value it guards, which is part of why supply-chain security was underinvested until this year.
| Device | List price (USD) | Secure element | Open source | Notes |
|---|---|---|---|---|
| Ledger Nano S Plus | $79 | ST33 (EAL5+) | No (BOLOS) | USB-C budget flagship |
| Ledger Nano Gen5 | $179 | ST33 (EAL6+) | No | E Ink touchscreen, BT and NFC |
| Ledger Flex / Stax | $249 / $399 | ST33 (EAL6+) | No | Premium touch; same security as cheaper Ledgers |
| Trezor Safe 3 | $59 | Optiga Trust M (EAL6+) | Yes | Entry secure-element model |
| Trezor Safe 5 | $129 | Optiga Trust M (EAL6+) | Yes | Color touch; the model CryptoBilis tampered |
| Trezor Safe 7 | $249 | TROPIC01 + Optiga | Yes | Openly auditable element; post-quantum firmware signing |
| Tangem (2-card) | $54.90 | Samsung (EAL6+) | App open | Seedless card, NFC, no screen, no firmware updates |
| Coldcard Q / Mk4 | ~$320 / ~$170 | Dual SE | Yes | Bitcoin-only, air-gapped, duress PINs |
| Keystone 3 Pro | $149 | Triple SE | Firmware open | QR-only air gap, broad asset support |
| BitBox02 Nova | ~$149 | Dual-chip | Yes | Swiss, compact, beginner-friendly |
| Foundation Passport | $199 | Secure element | Yes | Air-gapped, Bitcoin-only |
| NGRAVE Zero | ~$398 | EAL7 chip | Firmware closed | Highest chip grade, air-gapped |
| Blockstream Jade Plus | $149 | Virtual SE | Yes | No dedicated element; blind-oracle PIN |
Ledger: Sealed, Certified, and Carrying a Breach Ledger
Ledger remains the volume leader, with millions of devices sold and the most polished software stack in the category. The lineup runs from the $79 Nano S Plus through the $179 Nano Gen5, with its E Ink touchscreen and Bluetooth, up to the $249 Flex and $399 Stax. A reviewer should say plainly what Ledger will not: every one of these uses the same ST33-class secure element and the same closed-source BOLOS operating system, so the price difference buys polish, not security. On signing, Ledger has genuinely led, co-developing the clear-signing standard and handing its stewardship to the Ethereum Foundation in May 2026 so that approving a transaction shows you what you are approving.
The recurring problem is not the device; it is everything Ledger holds about you and everything that happens between its factory and your hand. A 2020 breach exposed roughly 270,000 customers’ physical addresses and has since produced a class-action suit seeking at least $500 million; a 2023 software supply-chain attack hit its Connect Kit; and in 2026 the CryptoBilis implants rode in through one of its authorized resellers. Chairman and chief executive Pascal Gauthier has made the company’s own case for why the device is the safe part, arguing on the Ledger blog that “everything that is in software is close to impossible to protect” and that “the safest place to store value is Bitcoin on a Ledger.” The device earns that confidence. The channel and the customer database are the parts a buyer has to manage, chiefly by purchasing direct and treating any Ledger-branded email as a probable phishing attempt.
Trezor: Open Source, and the Vendor That Shipped a Patch
Trezor’s pitch is transparency, and in 2026 that pitch paid off in public. The range is the $59 Safe 3, the $129 Safe 5 and the $249 Safe 7, the last pairing an Infineon element with the openly auditable TROPIC01. Chief executive Matej Žák frames the whole philosophy in a single line to DL News: “we are the most secure because we are the most open-source,” adding that in the past, with chips under non-disclosure agreements, “if we found a problem, we couldn’t warn anyone.” That is not marketing after this year. The CryptoBilis implant landed on a Safe 5, which is a mark against Trezor; but the open model is also why the attack could be dissected, confirmed in a lab, and patched in the open within days.
Grade the response on the channel-not-chip rubric and Trezor scores high: it treated an unverified customer report as credible, confirmed three devices in its own lab, shipped firmware 2.13.1 to neutralize the current implant batch, cut around 80 percent of its resellers, and offered replacements to every CryptoBilis buyer since March 2026. There is a real scar here beyond the implant: a third-party fulfillment partner’s breach exposed personal data on roughly 81,000 Trezor customers earlier in the year, the same data-is-the-weak-link theme that dogs Ledger. The buyer takeaway is that Trezor is the maker most likely to tell you the truth quickly when something goes wrong, which, after 2026, is itself a security feature.
Tangem: No Seed Phrase to Photograph
Tangem is the oddest device in the field and, by one specific measure, the most 2026-proof. It is a card, not a gadget with a screen: a Samsung secure element generates the keys on the card itself, you tap it to your phone over NFC to sign, and there is no firmware to update and no display at all. The headline property in the age of screen-reading implants is that, in its default mode, Tangem never shows you a 24-word recovery phrase, because it never generates one. As the company argued in its own note on the CryptoBilis incident, “there is no phrase to intercept if no phrase is generated,” and its app performs an authenticity check that flags a card activated before you set it up.
That strength is also the source of Tangem’s trade-offs, and an honest review names them. No firmware update means a structural flaw cannot be patched; the Donjon laser finding against the card is unfixable, though it needs physical possession and a six-figure rig to exploit. Seedless backup means you clone the key to two or three cards at setup and keep them apart, which is simple but offers no decoy wallet, no duress mechanism and no dead-man’s switch for inheritance. A two-card set runs $54.90 and a three-card set $69.90. Co-founder Andrey Kurennykh defends the non-updatable, seedless architecture as a deliberate choice rather than a compromise, and for a mobile-first saver who will never run an elaborate custody scheme, he has a point: the device structurally dodges the attack that defined the year.
The Air-Gapped and Bitcoin-Only Field
Beyond the big three sits a field of specialists, most built around an air gap: the device never connects by USB or radio, and signs transactions by passing QR codes or a microSD card. The reference Bitcoin-only signer is Coldcard, now the $170 Mk4 and the $320 Q, with dual secure elements, duress and brick PINs, and the most serious coercion-resistance features on the market. Coldcard also delivered the cautionary tale of the year on layer two: a five-year-old build flag weakened its random number generator and let thieves drain roughly $116 million from wallets whose seeds were generated with as little as 40 bits of entropy. The chip was fine; the firmware was not, and a later fix could not rescue seeds already created. That is the entropy question on the checklist, made concrete.
The rest of the field trades features against openness. Keystone 3 Pro ($149) is QR-only with three secure elements, broad asset support and open firmware. BitBox02 Nova (~$149) is a compact, open-source Swiss option that is friendly to beginners. Foundation Passport ($199) is an air-gapped, fully open, Bitcoin-only signer. NGRAVE Zero (~$398) carries the highest commercial chip grade but closed firmware. At the transparency extreme, Blockstream’s Jade Plus ($149) uses no dedicated secure element at all, splitting its secret across a self-hostable blind oracle, and the DIY SeedSigner is stateless, storing no key between sessions. One caveat ties the whole field back to CryptoBilis: an air gap closes the data channel, but the implant read the screen. A gap is necessary and not sufficient; it does not exempt an air-gapped device from the chain-of-custody question.
Security-Model Scorecard
The table below grades the models that matter against the layers 2026 actually tested: how the chip is governed, whether the code is open, whether your seed is ever shown on a screen, the incident that defined the year for each, and how the vendor behaved afterward.
| Brand | Secure element | Source model | Shows seed on screen? | 2026 incident | Vendor response |
|---|---|---|---|---|---|
| Ledger | ST33, sealed and certified | Closed (BOLOS) | Yes, at setup | CryptoBilis implant; legacy data breach, $500M suit | Direct sales unaffected; genuine-check gap acknowledged |
| Trezor | Optiga / TROPIC01 | Fully open | Yes, at setup | CryptoBilis Safe 5 implant; 81,000-record data breach | Firmware 2.13.1; ~80% reseller cut; free replacements |
| Tangem | Samsung, sealed on card | App open, no firmware updates | No, seedless by default | Donjon laser (lab only, no losses) | Unpatchable by design; app flags pre-activated cards |
| Coldcard | Dual secure element | Open | Yes, at setup | Entropy bug drained ~$116M | Firmware fix; cannot heal already-made seeds |
| Keystone | Triple secure element | Firmware open | Yes, at setup | None notable | QR-only air gap; tamper self-destruct |
| Jade / SeedSigner | Virtual / none | Fully open | Yes, at setup | None notable | Minimal attack surface; depends on user discipline |
Buying It Right: Chain of Custody Is Part of the Review
If the channel is now the attack surface, then how you buy is part of the device’s security rating. Mark Karpelès, the former Mt. Gox chief executive who has spent a decade studying how crypto gets stolen, put the CryptoBilis implant in blunt terms to BleepingComputer: “Ledger Nano X are used to store crypto. The spy implants are used to steal whatever crypto you store in those Ledger Nano X.” The device and the threat now travel in the same box. These are the buying rules that follow from that.
- Buy direct, or from a maker-verified store. Every confirmed CryptoBilis victim bought through a reseller or marketplace listing. The manufacturer’s own shop is the shortest, most controlled path.
- Initialize it yourself. A new device generates a fresh seed in front of you. A unit that arrives already set up, or ships with a card telling you to “use this recovery phrase,” is a scam; reject it on sight.
- Run the genuine check, but know its ceiling. Attestation proves the secure element is authentic. It does not detect a passive implant soldered beside the element, which is exactly how the CryptoBilis devices passed. Pair it with buying direct and a look at the packaging.
- Update firmware before funding. Trezor’s 2.13.1 only helps a device that installs it before it ever displays a seed. Patch first, generate the seed second, send coins third.
- If your reseller got cut, move. Anyone holding a device from a now-delisted channel should assume the seed may be known and sweep funds to a new seed on a direct-bought wallet.
What No Hardware Wallet Can Fix
Even a perfect device, bought perfectly, leaves three holes no firmware closes. The first is physical coercion. Chainalysis counted around $30 million in realized losses across 46 violent “wrench attack” incidents in the first half of 2026, with kidnappings in the majority of cases and France a hotspot; a separate CertiK tally, counting total exposure rather than realized loss, put the figure higher still. No device feature stops someone with a weapon, though duress PINs and decoy wallets buy time. Casa co-founder Jameson Lopp, who maintains a running database of physical attacks on Bitcoin holders, argues the only durable defense is privacy: do not advertise what you hold, especially under your real name or face. The same logic explains why thieves moved stolen CryptoBilis funds in seconds and why recovery is a race, the kind that starts the moment the theft is noticed.
The second hole is you at the keyboard. A hardware wallet signs what you tell it to sign, so phishing and malicious approvals route around the chip entirely. Data breaches at Ledger, Trezor and SafePal handed attackers exactly the names, emails and addresses needed to craft convincing lures, and a drained wallet looks identical whether the cause was an implant or a fake support site; telling an exploit from a con is its own hard problem. This is where clear signing earns its keep: if your device shows a human-readable summary of a transaction instead of an opaque payload, you have a chance to catch the theft before you approve it. The third hole is time: the same device that guards your keys so well can lock out your heirs, which is why estate planning, Shamir backups and multisig matter and why self-custody means you are the last line of recovery, with no help desk behind you.
That last point has a regulatory edge worth stating plainly for readers expecting a backstop. In the United States the Securities and Exchange Commission oversees exchanges, broker-dealers and token issuers, and its December 2025 investor bulletin on crypto asset custody draws the line clearly: the piece of hardware in your drawer is self-custody, and self-custody sits outside that perimeter. There is no chargeback on a drained hardware wallet and no regulator to petition. The autonomy is the product, and so is the risk.
The Verdict: Match the Signer to the Holder
There is no single best hardware wallet, and any review that crowns one is selling something. The right device is a function of who you are and what you are defending against. A first-time saver and a doxxed holder with a public seven-figure balance should not buy the same thing, and after 2026 both should think as hard about where the device comes from as about what is inside it. The matrix below maps common holders to a sensible starting point.
| You are | Best fit | Why |
|---|---|---|
| A first-time saver, modest balance | Tangem or Trezor Safe 5, bought direct | Seedless Tangem sidesteps the screen-reading attack; Safe 5 adds open source plus a vendor proven to respond |
| A multi-asset user across many chains | Ledger Nano Gen5 or Keystone 3 Pro | Broadest asset and app support with clear signing; buy direct and manage the phishing risk |
| A Bitcoin-only holder | Coldcard Q or Foundation Passport | Air-gapped, Bitcoin-focused, with the strongest duress and coercion features |
| A privacy or open-source purist | Trezor Safe 7 or SeedSigner | Open down to the chip, or a stateless do-it-yourself signer that stores no key at rest |
| A high-value or public holder | Multisig across brands plus collaborative custody | No single device or seed is a single point of failure; privacy and key separation beat any one gadget |
| Worried about supply chain above all | Tangem, or any device bought direct and patched first | Either show no seed on a screen, or control the channel and update firmware before funding |
Notice what the matrix does not say. It never recommends the most expensive device for its own sake, because within a brand the premium models buy finish, not safety. It leans on seedless designs and direct purchasing not because the chips are weak but because the chips were never the problem. And for the largest balances it stops recommending a single device at all, because the honest answer to concentrated risk is to split trust across keys, brands and people.
Whatever the Federal Reserve decides in December, and the market is already bracing for the hard part, the review criteria for this category have changed for good. The spy chip did not break any secure element, and the firmware fix that followed could not un-leak a single stolen seed. Both facts point the same direction. In 2026 you do not buy a chip; you buy a chain of custody, a vendor’s behavior under fire, and a design that limits what a tampered screen can even reveal. Grade those, and the chip will take care of itself.
Frequently Asked Questions
Are hardware wallets still safe after the 2026 implant attacks?
Yes, with a caveat. The secure elements in Ledger, Trezor and Tangem devices were never broken in 2026; the attacks went around them through the supply chain and user data. A hardware wallet bought directly from the manufacturer, initialized by you, and updated before it holds any funds remains the safest consumer option for self-custody.
What was the CryptoBilis hardware wallet hack?
CryptoBilis was an authorized reseller that shipped tampered Ledger and later Trezor devices carrying a hidden circuit board. The implant read the recovery phrase as the device displayed it and transmitted it over a cellular connection, letting thieves drain wallets; on-chain estimates of the losses range from about $86 million to more than $100 million.
Does Trezor firmware 2.13.1 make my wallet safe again?
It helps, but it is not a full reset. Firmware 2.13.1, released on 11 November 2026, stops the known implant from capturing data on any device that installs it, but it cannot protect a recovery phrase that was already captured. Anyone who bought a Trezor from CryptoBilis should move funds to a new wallet with a new recovery phrase rather than rely on the patch alone.
Which hardware wallet is best in 2026?
There is no single best device; the right choice depends on your threat model. Tangem suits savers worried about supply-chain tampering because it never shows a seed phrase, Ledger and Keystone suit multi-asset users, Coldcard suits Bitcoin-only holders, and anyone guarding a large or public balance should consider multisig across more than one brand.
Is Ledger or Trezor more secure?
Both rely on strong, certified secure elements, so neither is insecure at the chip level. Trezor is fully open source and handled the 2026 implant discovery with a fast public fix and an 80 percent reseller cut, while Ledger offers the most polished software but carries a heavier history of customer-data breaches. Whichever you choose, buy it directly from the maker.
Yuki Tanaka is a senior wallets and self-custody writer at HOGE Wire, covering hardware security, key management and the economics of holding your own coins.