Ledger’s Ethereum App Bug: What the APDU Race Condition Meant for Users

Ledger’s Ethereum app contained a flaw in certain clear-signing flows that could allow a malicious dApp to substitute transaction data while a user was reviewing the original transaction. In the feared scenario, the Ledger screen would continue showing legitimate details even though the device signed a different request. 39

Ledger says its Donjon security team identified and fixed the issue before TestMachine’s public disclosure. The patch shipped in Ethereum app version 1.22.2 on August 12, 2026, while TestMachine began publishing its findings on August 21. 569

How the Ledger signature-substitution attack worked

The issue involved the handling of APDU commands—the messages exchanged between a connected host or dApp and the Ledger Ethereum app. During a normal clear-signing flow, the device receives transaction data, presents important details for review, and waits for the user’s approval.

According to the reported research, a malicious website or dApp with WebHID access could send a second APDU command while the first transaction was still under review. That created a race between competing signing requests. The vulnerable flow did not reliably bind the transaction shown on the device to one immutable signing session, creating a path for signature substitution. 3715

The practical danger was not simply that a transaction could fail. A user could see an innocuous action and approve it while the device ultimately signed a malicious token approval, transfer, or other substituted request. That would undermine the central purpose of clear signing: checking transaction details on the hardware device before authorizing them. 110

What version 1.22.2 changed

Ledger’s Ethereum app update addressed the vulnerable signing flow. Technical reporting on the code changes says version 1.22.2 prevents a new signing session from replacing one already under review and rejects an approval callback when the application state no longer matches the active request. 10

Ledger’s CTO, Charles Guillemet, said the Donjon team found the vulnerability with AI-assisted vulnerability-detection tools and deployed the fix on August 12. Reports described the accompanying release-note language as a brief security-fix notice rather than a detailed public advisory. 356

That distinction matters for users. Installing a new app version can protect the signing path going forward, but a minimal changelog gives device owners little information about what changed or whether they needed to take immediate action.

TestMachine and Azimuth’s public disclosure

TestMachine said its AI agent, Azimuth, found the issue during an autonomous scan and that the team validated the behavior on a Ledger Flex. In posts published from August 21 to August 23, TestMachine described how a malicious dApp could race an APDU command during transaction review and framed the issue as affecting every Ledger running the Ethereum app. 4715

The public disclosure brought attention to a vulnerability that Ledger says had already been patched. TestMachine’s account also said the company declined a bounty, while Ledger disputed how the communication unfolded. 1612

The dispute over who found it first

The disagreement is primarily about the private discovery and disclosure timeline—not the existence of the patch.

Ledger’s position, stated by Guillemet, is that Donjon discovered the flaw, fixed it, and shipped the correction roughly two weeks before TestMachine went public. He criticized the subsequent disclosure as creating unnecessary panic and said the company had already addressed the vulnerability. 6712

TestMachine’s position is that Azimuth independently found and validated the issue, and that Ledger’s quiet release did not meaningfully warn users about the risk. The company’s public posts emphasized the attack scenario and the breadth of its claim. 715

The available reporting supports the August 12 patch date and the August 21–23 public-post dates. It does not independently establish the exact private discovery dates, disclosure communications, or whether either account provides the complete chronology. Those details should therefore be treated as competing statements rather than settled facts. 56912

Which Ledger devices may have been affected?

TestMachine’s broad claim was based on shared Ethereum application and signing code, but its reported hands-on validation was performed on a Ledger Flex. Reports identified modern device families that may share relevant code, including the Nano S Plus, Nano X, Stax, and Flex. 2720

That does not amount to a complete, independently documented proof of concept across every Ledger model. As of August 24, the phrase “every Ledger” remained a researcher claim rather than a fully demonstrated cross-device result. Users should distinguish between a potentially shared software path and a publicly reproduced exploit on each individual device line.

Were any users’ funds stolen?

As of August 24, 2026, no independently verified thefts tied specifically to this vulnerability had been reported. There was also no confirmed complete public demonstration covering every claimed Ledger device. 25614

That is a statement about the evidence available at that time, not proof that the flaw was never exploited. The vulnerability was serious because it could have defeated the transaction review users rely on, even though confirmed losses had not emerged in the available reporting.

What Ledger users should do

Users should open Ledger Live and update the Ethereum app to version 1.22.2 or later. They should also keep Ledger firmware and installed applications current. The relevant remediation for this signing flaw was the Ethereum app update—not merely a firmware update. 515

After updating, users should continue treating on-device review as essential: verify the recipient, amount, and contract action shown on the Ledger before approving any transaction. The update addresses the reported session-substitution path, but careful transaction review remains an important security practice.