What Is Blind Signing? Why Your Hardware Wallet Can't Show What You Approve in 2026
Blind signing is the last step in most crypto thefts: a human approves a transaction the device cannot explain. Two 2026 launches finally address it — and coverage, not the standard, is now the bottleneck.
What Is Blind Signing?
Blind signing is approving a transaction your wallet cannot display in plain language. The device knows it is signing something. It cannot tell you what. Instead of “swap 1,000 USDC for 0.42 WETH”, the screen shows a hash, a checksum, or a block of hexadecimal.
The cause is mechanical. Your computer builds the transaction as calldata — a function selector plus its arguments — and hands the device only that. A router swap is a wall of 32-byte words padded with zeros; nothing in it says “you are sending 1,000 USDC.” The readable sentence you saw was rendered by a web page, not by anything you trust.
So the security property inverts. The private key stays offline, but verification — the only reason a hardware wallet exists — quietly moves back onto the machine that may already be compromised.
Why “Check Your Hardware Wallet” Stopped Working
For a plain transfer the advice holds: the device decodes the recipient and amount itself and shows both in text. Contract calls are different. There the small screen asks you to confirm a value no human can evaluate, so the confirmation becomes faith in the laptop.
Wallet drainers and compromised front ends are built around that gap: the malicious page renders a harmless description while the calldata does something else. Nobody needs your seed phrase when you will sign the payload yourself.
The tell is simple. If the device shows a hex string and the friendly description exists only in a browser tab, you are trusting a screen the attacker may control.
Two Cases That Define the Risk
Radiant Capital, October 2024. Attackers planted malware on contributors’ machines and waited. Signers approved what looked like a routine transaction and handed over ownership of the lending contracts. Around $50 million left.
Bybit, February 2025. Attackers compromised the Safe{Wallet} front end and served modified interface code. The transaction was a delegateCall that replaced the safe’s implementation. The screens said one thing; the payload did another. Roughly $1.4 billion — still the largest crypto theft on record, attributed to the Lazarus Group.
Both had audited contracts, cold storage, multiple signers and separated keys. None of it mattered, because the last human in the loop approved data they could not read.
What Changed in 2026: ERC-7730 Clear Signing
On 12 May 2026 an Ethereum working group of wallet developers and security firms, with the Foundation’s Trillion Dollar Security Initiative, launched Clear Signing as the structural answer. The mechanism is a JSON descriptor format, ERC-7730, initiated by Ledger and now neutrally stewarded by the Foundation.
A descriptor maps a contract function or EIP-712 message to human fields: an intent, a token amount with correct decimals and ticker, a labelled recipient, the protocol name. Wallets fetch the descriptor at signing time and render it, so the trusted screen — not the browser — becomes the final authority.
Descriptors are published to an open, permissionless registry at clearsigning.org. Entries carry attestation identifiers so review signals can be resolved, but the trust decision stays wallet-local: publication does not imply display, and each wallet sets its own policy. The system is deliberately rebuildable — if today’s operators vanished, anyone could reconstruct the registry from published files.
Trezor Shipped It. Coverage Is the Story.
Trezor enabled Clear Signing in firmware on 7 September 2026 for the Safe 7, Safe 5, Safe 3 and Model T — but not the Model One, a 2014 device still sitting in plenty of drawers. There is no switch to flip: supported transactions render in plain language through Trezor Suite, WalletConnect and Trezor Connect. Launch coverage named 1inch, Aave, Lido, Tether, LiFi and Hyperliquid.
The number that matters is not the partner list. Of the twenty largest Ethereum applications by deposited capital, exactly two were on the plain-text list when the feature went live. Published descriptors had grown about 28% between May and the end of July 2026 — real growth from a small base.
Anything without a descriptor falls back to the old flow, now with a warning. In 2026 that fallback is where most users still are. Clear Signing fixed the mechanism; it has not fixed the coverage.
Blind Signing Is Not Only Calldata
The same failure lives in typed-data signatures. A permit grants a token allowance with no transaction and no gas, arriving as a structured EIP-712 message — and most wallets render those fields raw. A setApprovalForAll signature is worse: one approval, every NFT in the collection.
ERC-7730 covers EIP-712 messages too, which matters because “sign this message” prompts are where the least readable data meets the fastest approvals. A signature is not a cheaper transaction. In a permit, it is the transaction.
What to Do Before You Approve
- Read the device, not the browser. A hash on the trusted screen means cancel and treat the prompt as hostile.
- Check the spender and the amount. Unlimited approvals are a standing liability, not a convenience.
- Run current firmware. Clear Signing arrives by firmware update; an old device is blind by default.
- Ask what is actually covered. Descriptor support is per contract, not per wallet.
- Simulate, then verify. Simulation shows outcomes, not intent — a drainer can simulate cleanly.
- Separate hot from cold. Long-term holdings should never see a contract approval.
- Audit standing approvals. A signature made two years ago is still executable today.
Institutions need a policy, not a device: an allowlist of descriptor-verified contracts, escalation to a second reviewer for any blind-signature request, and logs of what each signer actually saw.
Blind signing survived this long because it was invisible — a failure that felt like normal operation. Clear Signing makes it visible. The rest is discipline.
Frequently Asked Questions
What is blind signing in crypto?
Blind signing is approving a transaction whose contents your wallet cannot display in readable form. The device signs a hash or a block of hex calldata while the human-readable summary exists only on the computer, which may be compromised.
Did clear signing fix blind signing in 2026?
No, it made it optional rather than universal. ERC-7730 Clear Signing launched in May 2026 and Trezor shipped firmware support in September 2026, but at launch only two of the twenty largest Ethereum applications by deposited capital had a descriptor, so anything uncovered still falls back to blind signing with a warning.
What is ERC-7730?
ERC-7730 is an open Ethereum standard that describes smart contract calls and EIP-712 messages as human-readable JSON descriptors. Wallets fetch the descriptor at signing time and show intents, amounts, tickers and labelled addresses on the trusted device screen instead of raw calldata.
Does a hardware wallet protect you from blind signing?
It protects the key, not the decision. If the device can only show a hash, then a compromised computer controls what you believe you are approving, which is exactly how Radiant Capital lost about $50 million and Bybit lost roughly $1.4 billion.
How can you tell you are about to sign blind?
If the trusted screen shows a hex string, a checksum, or a “data preset” prompt instead of an action, amount and recipient, you are signing blind. Cancel, verify the contract through a second source, and treat any urgency around that prompt as a red flag.
Know what your stack is asking users to sign
Blind signing is a product decision, not just a user mistake: undocumented contract calls, unverified implementations and stale approvals are all visible from the outside. RootCrak’s autonomous scanner probes your domains, APIs and Web3 infrastructure and surfaces the gaps before someone signs them away.
Get a Free Security Scan