Trezor Safe 5 implant report via CryptoBilis: device unexamined, Trezor says 4 inspected showed no tampering

One unexamined report, and four inspected devices with no tampering found

Trezor says its security team received a responsible disclosure on 28 October 2026 reporting a hardware implant in one Trezor Safe 5 bought from the reseller CryptoBilis in August, and that it published its warning two days later (derived: 28 to 30 October). It has not seen the device: "We haven't received the device to confirm this report". Yet it asks everyone who bought a Trezor from CryptoBilis from March 2026 onwards and is using it to store funds to move them to a new wallet with a new seed. Its own timeline graphic adds a fact that the post's text does not carry: four other devices bought from CryptoBilis were returned for inspection from 19 October, and "Our inspection found no evidence of tampering." That is one unexamined report against four inspected devices with no finding (derived).

What that does not establish is that the implant exists. Trezor's own words are "a responsible disclosure report alleging a hardware implant", "treated as a credible report" and, of the report's description of what the board does, "we have not verified this ourselves". It does not establish how many devices are affected, who made or fitted anything, or whether any funds were taken. Trezor says none has been reported stolen, but it also says it cannot see what its customers hold. The warning is a precaution, and Trezor says so: waiting for lab confirmation "would leave CryptoBilis buyers exposed in the meantime".

This briefing reads Trezor's own post, Implant reported in a Trezor sold by CryptoBilis, dated 30 October 2026, in full from 13:58 UTC that day, and the Singapore Police Force advisory of 28 October. The post's timeline section is a graphic, so its text was read from the image. The Hacker News blocks scripts and was read by a person in an ordinary browser tab; it is a secondary source and is used only to compare. No account was created and no form was filled in. Trezor's robots.txt and the Singapore Police Force's were each read first and do not bar the pages used.

Trezor sells the wallets and wrote the only account of the report, and a maker has an interest in being seen to act early and in placing the fault with a reseller; the warning also costs it a replacement device or a voucher for every eligible buyer, so it is not free. The report itself is one customer's, relayed by the maker, and nobody outside Trezor has examined the device.

What Trezor says it found, how it knows, and what it leaves out

The table sets Trezor's statements against what neither its post nor its timeline states. "Not stated" means not found in the post's text or its graphic, both read in full. It does not mean the information does not exist.

Trezor's post is not consistent with itself on one point. It says the implant "appears to operate similarly to those seen in Ledger devices", and then, in its FAQ, "we won't draw comparisons while investigations are ongoing". Read together, Trezor is saying the description resembles the Ledger cases and declining to say more. The Hacker News reports the same two points. The Ledger side, as far as Ledger has stated it, is in this site's earlier briefing on the Ledger reseller losses, which was written from Ledger's own posts of 9 to 11 October and does not cover Trezor. Trezor's timeline says Ledger's Donjon team published details on 13 October of an implant inside the device that "reads what the screen shows and sends the recovery phrase out over LTE". That is Trezor's summary of Ledger's publication; Ledger's own post was not read for this briefing.

The window is open at one end. Trezor says "from March 2026 onwards" and gives no reason or end date; it removed CryptoBilis on 9 October. From 1 March, the earliest the wording allows, to 9 October is at most 222 days (derived). Its text and its timeline also describe the 9 October step differently: the text says Trezor "removed them", the timeline says Trezor "asks CryptoBilis to stop selling all Trezor devices. CryptoBilis complies."

Trezor's own timeline: four inspected devices and 19 days

The graphic below is Trezor's own, reproduced as published. Its entries run from 9 October, when Ledger opened its investigation and Trezor asked CryptoBilis to stop selling, to 28 October, when the customer's report arrived. Two entries are the ones the post's text omits: on 14 October Trezor "asks customers who bought a Trezor from CryptoBilis to send their devices to us for inspection", and from 19 October "Four devices were returned: a Trezor Safe 7 ordered on April 6th, two Trezor Safe 5s ordered on June 7th, and a Trezor Safe 3 ordered on August 4th."

Zoom

The arithmetic, all derived from Trezor's dates: Trezor asked buyers to send devices on 14 October and the report came on 28 October, 14 days later; Trezor's removal of CryptoBilis on 9 October was 19 days before it. Of the four devices Trezor says it inspected, two were Safe 5s, the model in the report, both ordered on 7 June, and none was ordered later than 4 August. The reported device was bought "in August"; the post gives no day.

What do four devices with no finding show? Very little on their own, and the arithmetic says why. If a tampered share of CryptoBilis devices were present and four were drawn at random, finding none would be unremarkable at any share up to about 53 per cent: the exact 95 per cent upper bound on the share after 0 findings in 4 is 52.7 per cent (derived: 1 minus 0.05 to the power of one quarter). The four were not drawn at random. They are the devices whose owners chose to send them in, the post does not say how many were asked or how many sold, and "onwards" leaves the number returned open. So the four are a fact worth reporting, because the post's text leaves them out, but they neither clear the channel nor undercut the one report. The post does not say how the inspection was done, whether devices were opened, or what Trezor would have looked for.

Zoom

What the report describes, and what a defender can and cannot infer

Zoom

The reported mechanism is simple to state and Trezor has not confirmed it. If it is true, the board does not need to break the wallet's cryptography. It reads the screen, and the screen is where a Trezor shows the recovery phrase to its owner (the "wallet backup" in Trezor's words). That is this briefing's reading, not Trezor's wording. It follows from Trezor's description that anything shown on the screen while such a board is fitted would be exposed, which fits Trezor's advice to move to a new wallet with a new seed and its statement that using the backup generated by the tampered wallet on another device does not solve the issue.

One inference is worth stating plainly, and it is an inference. Trezor's own pages describe its authenticity check as a certificate stored in the Secure Element, and its bootloader as checking the firmware signature each time the device is connected (device authentication, authenticate the Safe 5, read 30 October). Neither page says that the check would detect an extra board sitting between the main board and the screen, and the post does not say whether the reported device passed it. For its own wallets Ledger says in plain words that a genuine check "cannot detect unauthorized physical modifications to the hardware, such as spying implants, if the original Secure Element remains intact" (Ledger Donjon, read 30 October). That is Ledger's statement about Ledger's check, and Trezor has not made a matching one. The point is narrow: a pass on the maker's software check is not evidence about a board added beside it.

Nor does the post say whether the reported device's seal was intact, replaced or missing, whether it arrived with firmware already installed, or whether its backup cards were blank. Those are exactly the checks Trezor's guide tells a new owner to make, and their results are the first things a reader would want to know.

Four friendly names: authorised reseller, seal, authenticity check, no reports

A comforting label is not a control. Each of these names does a job in a buyer's head, and the record shows what each one covers and where it stops.

On the first row the point is not that the label was dishonest. It is that "authorized" described a commercial relationship, and nothing in the text read says it was also a check on what a reseller did to a device before it reached a buyer. Trezor's resellers page lists its authorised resellers by country, and its safety page says "Never buy a Trezor from an unauthorized third party, as you can never tell who may have had access to it before you." The incident Trezor reports would, if confirmed, be exactly that worry arising inside the authorised channel.

A separate Singapore police advisory, about Ledger and not Trezor

The Hacker News adds the Singapore Police Force advisory of 28 October 2026, and it is a different case. The advisory, read in full, says "Since 10 September 2026, at least two cases have been reported, with total losses amounting to about $3.5 million". The advisory does not say which currency the dollars are. It says Ledger wallets "purchased from some retail stores in Singapore contained malicious hardware components" that could capture recovery phrases or private keys, that the police are in contact with Ledger, which is reviewing the cases, and that investigations are ongoing. It names Ledger only. It does not mention Trezor or CryptoBilis, and it does not say whether the stores were authorised resellers.

So it is not evidence about the Trezor report. It is a second regional case in which police say, on preliminary investigation, that wallets bought in shops carried malicious hardware, and it gives owners advice that agrees with Trezor's: refer to the maker's own advisories, avoid holding funds on a suspect device, and move remaining funds to a new wallet with a newly generated recovery phrase on an uncompromised device. The Hacker News's article leaves out the $3.5 million figure.

Where the press differs from the primary

Technical details

Trezor's post prints no indicators of any kind: no hashes, serial numbers, part names, firmware versions, domains, addresses or photographs in its text. There is nothing to hunt for on a network, and the described board uses a mobile network, which would not touch the buyer's own. That second point is this briefing's inference from Trezor's description. What a defender has is the text of the report and the owner checks Trezor publishes for the Safe 5. Both are copied below exactly as printed, with the source and date read in each caption.

The first block is Trezor's description of what the report says was found. It is an allegation relayed by the maker, and it is the only technical description published.

The report describes an extra circuit board fitted inside a Trezor Safe 5, between the main board and the screen, with a cellular module and an antenna. According to the report (we have not verified this ourselves), it reads what the screen shows, including the wallet backup, and sends it out over a mobile network whenever the device is plugged in.

The next block is Trezor's own guidance for what a new owner of a Safe 5 should check before setup: the seal, the backup cards, the firmware and the authenticity checks. These are the steps to run on any new device, and a failed step is a reason to stop and contact the maker. The guide is written for counterfeits and damaged packaging. Whether it would find the board in this report is not stated.

Your Trezor Safe 5 comes with a holographic seal over the USB-C connector. The seal should be clean, firmly attached, and free from tears or residue. If the seal was damaged or missing when you received the device, pause setup and contact Trezor Support for guidance.
Your package includes two blank wallet backup cards (20 words). These cards should be completely empty. You fill them in yourself during setup when your Trezor generates a new wallet. If your wallet backup cards have words already written or printed on them, do not use the device and contact Trezor support.
Every Trezor Safe 5 leaves the factory without firmware installed. During setup, Trezor Suite guides you to install the latest signed firmware.
Trezor Safe 5 uses a Secure Element (OPTIGA Trust M) to verify that your device is genuine and unmodified.
Each time you connect your Trezor, the bootloader checks the firmware signature. Trezor Suite will only work with firmware officially signed by SatoshiLabs. If the signature is invalid, your device and Suite will warn you.

The last block is Trezor's scope text for the warning, which is what an organisation should match its purchase records against.

IMPORTANT: If you bought a Trezor from CryptoBilis from March 2026 onwards and are using it to store funds, please move your funds to a newly created wallet with a new seed as soon as possible.
Any store selling devices supplied by CryptoBilis, including its store on Tokopedia. If you’re not sure where your device came from, check the seller name on your receipt or contact Support.

What to hunt and harden, from these artifacts: the seller name and date on every hardware wallet purchase, the seller of any device that came by a marketplace, whether the device arrived with firmware already on it or with words on its backup cards, and which devices have already shown a recovery phrase on screen. Nothing printed lets a defender test a device remotely, and Trezor says an owner should not weigh or open it.

What owners and organisations should do, in order

Zoom

The question the record leaves open

Trezor moved from a report to a public warning in two days, on a description, photos, a receipt and a video call. That is a defensible call, and the post argues it well. It leaves open a plainer problem. For up to about seven months (derived from Trezor's own window), "authorized" was the word between a buyer and a device that passed through a reseller's hands, and the one inspection result Trezor offers, four devices with no tampering found, was gathered only after Ledger's investigation began on 9 October. The device that could settle it has not arrived.

If the word "authorized" and an intact seal were the only controls between the maker's factory and the buyer's desk, what check were they standing in for?

Key facts