BitBox02 Review: Two Chips, Two Editions, Three Fixes

The BitBox02 is a Swiss made hardware wallet from Shift Crypto, built around two chips and sold in Multi and Bitcoin only editions that cannot be swapped later. Its August 2026 Dixence update fixed two severe firmware flaws, and the vendor’s own advisory says the Bitcoin only edition escaped one of them.

On this page: the chips named in BitBox’s own datasheets and security page, the three issues disclosed in the 17 August 2026 Dixence advisory with the exact firmware ranges each one hit, prices read off the vendor’s shop on 5 October 2026, and five dated accounts from Trustpilot and the public issue tracker.

What is the BitBox02, and which editions are on sale?

Two product lines carry the name today. The original BitBox02 connects by USB-C only, and the BitBox02 Nova adds a glass display, an optional Bluetooth path and support for iPhone and iPad, which the original does not. Each comes in a Multi edition and a Bitcoin only edition. Read off the vendor’s own shop on 5 October 2026, the BitBox02 was listed at 124.17 euros and the Nova at 145.83 euros, both excluding VAT, each with a two year warranty.

The sentence that matters most on those pages is the dullest one. BitBox writes that the editions “are two different products that can’t be converted later”, so the choice is made at checkout and not in software. Multi adds Ethereum and ERC20 tokens, Cardano, Litecoin, what the shop describes as 1500 or more further tokens, and FIDO U2F login. The Bitcoin only edition strips all of that out, which BitBox frames as removing attack surface rather than features.

How does the dual chip design actually protect the seed?

Every hardware wallet page says “secure chip” and stops there. The useful detail is which chip, and what it is allowed to know. The firmware repository names the microcontroller as a Microchip ATSAMD51J20A, a 120 MHz Cortex-M4F, talking to a separate secure element over I2C. The security page, read on 5 October 2026, names that element: ATECC608B in the BitBox02, OPTIGA Trust M V3 in the Nova.

The split is the point. BitBox states that three secrets are needed to reach the encrypted wallet seed: one held on the secure chip, one on the microcontroller, and the device password, which lives only in the owner’s head. The seed stays encrypted in RAM while the device is in use and is decrypted only for as long as a signature takes. A monotonic counter in the secure chip caps password attempts and password stretching makes each guess slow, which is how a short password survives contact with a patient thief.

The Nova datasheet adds the parts a buyer can check: five sources of entropy for seed generation, an attestation key written at the factory so the app can spot a counterfeit, a bootloader that refuses unsigned firmware, downgrades and firmware built for the other edition, and the option to display the firmware hash before it runs. Both chips sit under security grade epoxy, and the casing pins break if the halves are prised apart. None of that is exotic in 2026. All of it being published still is not standard. The firmware is reproducible and WalletScrutiny tests that claim independently, a subject the guide to open source crypto wallets goes into properly.

What did the August Dixence update fix, and on which editions?

On 17 August 2026 BitBox published the Dixence update and firmware 9.26.5, disclosing three security issues. The advisory says they were found during internal audits, including with what it calls frontier AI models, that there are no reports of stolen user funds, and that “Your existing wallet seed is not affected in any way”. Bitcoin Magazine covered it the next day. This is the part no affiliate review carries, because a security notice sells nothing.

The three issues, as the vendor describes them. A bootloader flaw that could have been used to walk an owner into installing malicious firmware on a genuine device, after a successful phishing step; it was fixed earlier, in the Oeschinen release, and the Nova was never affected because of the age of the bootloader versions involved. A memory corruption issue that applies only to a Multi edition with no wallet set up on it yet, plugged into a malicious host, where it could allow arbitrary code execution. And a silent payments flaw that could have locked coins to an unintended address, a ransom scenario rather than a theft. The affected ranges are specific, so here they are as the advisory states them, with the fixing release and its date from the public release record.

Two footnotes worth having. BitBox credits external researcher Jan Wütherich of SySS GmbH with independently reporting the bootloader issue after its own engineers had found it. Separately, the trade press credited a CertiK researcher with an out of bounds write in the USB path; the release notes for firmware 9.26.3, dated 8 July 2026, do record “Fixed potential out of bounds write”, but the Dixence advisory does not name CertiK, so treat that attribution as reported rather than vendor confirmed.

Does the Bitcoin only edition really have less attack surface?

“Less code means less attack surface” is the kind of line that usually means nothing. In August 2026 it briefly meant something measurable. The memory corruption issue did not apply to the Bitcoin only edition, and the reason BitBox gives is refreshingly literal: “its firmware does not contain the affected code”. KeepKey is only now heading the same way, and its first Bitcoin only build is still a pre release, as our check on whether KeepKey firmware is still maintained sets out.

That is one issue out of three, though, and the honest reading cuts both ways. A Bitcoin only BitBox02 was still inside the affected range for the bootloader issue and for the silent payments issue. Counting the three disclosures across the four products shows how the editions differ, and it is a crude count on purpose: an issue tally is not a severity ranking, and one bootloader flaw matters more than two that need a hostile computer and a device nobody has set up.

What do BitBox02 owners actually report?

We read about thirty seven Trustpilot reviews of bitbox.swiss and one long thread on the public issue tracker, on 5 October 2026. The headline numbers: 4.8 out of 5 across 2,162 reviews, one star reviews at 2 per cent, two star reviews under 1 per cent. These are accounts of what people say happened to them, not findings, and most of the positive ones praise the support team rather than the device, which is what a solicited review pile tends to look like. Five of them, dated.

• A two star reviewer in May 2024 reported the touch slider misreading their input badly enough to burn four of their ten password attempts. A one star reviewer in May 2023 described the touch edges as too sensitive to use, and another in October 2024 reported the touch interface malfunctioning.
• A one star reviewer in February 2025 reported a 69 dollar UPS brokerage fee landing after a 129 dollar purchase, taking the total to roughly 200 dollars. Another in December 2025 put theirs at about 250 dollars. Both were shipping into the Americas.
• A one star reviewer in July 2024 reported the device rejecting their password after they had lost the microSD card, and losing access as a result. That is the backup story every custody guide tells, arriving in the post.
• Reviewers in September 2026, rating four and five stars, still flagged entering a long passphrase on the device as the awkward part, alongside praise for the hardware and the support.
• On the BitBoxApp tracker, a contributor opened issue 4053 on 24 April 2026 arguing that the app records the existence of passphrase protected wallets in its accounts file and its log, which undercuts a hidden wallet if the computer is later examined. Maintainers replied through 30 April that the app cannot currently tell a passphrase keystore apart without device side changes, that the upkeep would be permanent, that support staff rely on the logs, and that an incognito mode would be the cleaner fix. The issue is open and low priority.

Where the accounts agree with the documentation: the microSD backup is the single point of recovery, and the people who lose access are the people who lost the card. Where they add something it does not say: the touch slider is the part that goes wrong in the field, and on a device with a documented ten attempt limit a slider that misreads a gesture is a custody risk rather than an annoyance. Import duty is the other one, nobody’s fault and real money all the same. If a long passphrase is part of the plan, the guide to what a recovery phrase is and how to store it is worth reading first.

How does the BitBox02 compare with the Nova and the alternatives?

Prices were read from the vendors’ own stores on 5 October 2026. The chip and connectivity rows come from vendor datasheets and security pages, not from review sites.

The Nova’s Bluetooth deserves a sentence, because Bluetooth on a signing device sounds alarming. BitBox calls the implementation Whisper and describes it as a minimal trust architecture: end to end encrypted, running on a separate chip with no memory and no access to wallet secrets, and able to be disabled entirely. That is a vendor design claim rather than an audited finding, and the owner can switch it off. If the whole category is new, start with what a cold wallet is, and if an air gap matters more than convenience the Coldcard Q takes a different route again.

What does the BitBox02 security page leave unanswered?

For a page this detailed, the audit line is remarkably thin. In full, it says the BitBox02 firmware “was audited by Census Labs along with consulting done by multiple third-party security firms”. No date, no scope, no published report to read, and we could not find one. Every ranking page repeats “independently audited” without noticing that the claim, as written, cannot be checked by a reader. The same trick runs through certification badges, which is why we went looking for the document behind the NGRAVE ZERO EAL7 claim and found it covers an operating system rather than a wallet.

The second gap is where disclosures live. Checked on 5 October 2026, the firmware repository has no published GitHub security advisories, so there is no CVE feed to subscribe to and no notification unless you read the blog or run the app that prompts you. BitBox also wrote that it “may share more technical information in a writeup at a later time”, which had not appeared when we checked. Set against the published threat model, the schematics, the x-rays and the named chips, those are small gaps. They are still the two a careful owner would want closed.

The practical conclusion is the boring one, and it holds for every brand covered here: the firmware version on the device is the security feature. BitBox says it will never ask for the recovery words, and nothing legitimate ever will, so an update prompt arriving by email rather than inside the app is a phishing attempt and not a favour.

Frequently asked questions

This article is educational and is not financial advice. Nothing here is a recommendation to buy, sell or hold anything. You are responsible for your own security decisions.