# Coldcard seed entropy warning after 594 BTC drain > Coinkite warned Coldcard users after ~594 BTC was swept from ~500 single-sig wallets. Mk3 seeds from firmware 4.0.1+ are critically weak; newer models need the July 31 hotfix. - **Source:** https://ptycoin.com/en/posts/2026-07-31-coldcard-mk3-seed-entropy-warning-594-btc/ - **Published:** 2026-07-31 - **Category:** News - **Author:** Mateo - **Tags:** bitcoin, security, self-custody, wallets - **Also published in:** [Español](https://ptycoin.com/es/posts/2026-07-31-coldcard-mk3-seed-entropy-warning-594-btc/) --- **Coinkite** published a [Coldcard security advisory](https://blog.coinkite.com/coldcard-mk3-seed-generation-warning/) on **30 July 2026** after on-chain investigators reported roughly **594 bitcoin** swept from about **500** single-signature addresses in a short run of blocks. The Canadian hardware maker says seeds generated on affected firmware may have far less randomness than users were told, and it shipped [hotfix firmware 5.6.0 and 1.5.0Q](https://coldcard.com/docs/upgrade/) on **31 July**. [The Block](https://www.theblock.co/post/410235/coinkite-coldcard-mk3-warning) and [CoinDesk](https://www.coindesk.com/tech/2026/07/31/major-bitcoin-wallet-flaw-drains-594-btc-in-25-minute-sweep) matched the same core sequence: a coordinated drain, then Coinkite’s public warning. Coinkite has not published a closed-form forensic proof that *this* sweep was caused solely by the entropy bug, and its investigation is ongoing. Security researchers examining the chains of spends have linked the pattern to weak seed generation on Coldcard devices. ## What moved on-chain According to [Atlas 21’s reconstruction](https://atlas21.com/594-bitcoin-drained-15-minutes-theft/) cited by The Block, about **594.5 BTC** (near **$38 million** at then-prevailing prices) left roughly **500** single-signature wallets across Bitcoin blocks **960188** to **960191** on **30 July**. Many of the source addresses had been dormant for years and held more than **0.15 BTC** at the time of the spend. That is the public “known set” most outlets used for the first wave. [Clay Garrett of Block’s Bitcoin engineering and security team](https://x.com/clay_garrett/status/2082980439367487724) later flagged **695** earlier transactions that shared the same full fingerprint as the known set, moving another **~488 BTC**. If those earlier spends are the same operation, the combined total approaches **1,083 BTC**. That second tranche is still an investigator’s extension of the fingerprint cluster, not a court-verified total. ## What Coinkite says is broken Coldcard is an air-gapped, Bitcoin-only hardware signer from Coinkite: the private keys stay on the device, and you sign transactions without exposing the seed to a networked computer. The product’s security story has always depended on how the device builds that seed (the BIP-39 mnemonic) when you choose “new wallet.” Coinkite’s advisory is blunt about the entropy shortfall: - **Mk3:** seeds generated on firmware **4.0.1** (March 2021) through **5.0.3** (the last Mk3 release) may have only about [**~40 bits**](https://blog.coinkite.com/entropy-technical-backgrounder/) of entropy. That is critically low. Coinkite says it is **exploring one final Mk3 release** but warns that updating the deprecated platform risks bricking units; either way it tells those users to migrate. - **Mk4, Mk5, and Q:** seeds generated **before** fixed firmware may have as little as about **~72 bits**, still well below the **128-bit** target. Fixed builds are **Mk4/Mk5 5.6.0+** and **Q 1.5.0Q+**, released **31 July 2026**. - **Not affected:** [TAPSIGNER, OPENDIME, and SATSCARD](https://blog.coinkite.com/coldcard-mk3-seed-generation-warning/) use different codebases and are outside this bug. Early media notes that said “only Mk3” tracked Coinkite’s first public framing. The live advisory and the upgrade page now put newer models in scope until they run the hotfix and you generate a *new* seed on that fixed firmware. An update **cannot** repair a weak seed that was already written down years ago. ## Who is actually exposed The risk is **not** “every Coldcard on a desk.” It is seeds whose randomness came from the broken device generator, used as single-signature wallets without extra entropy. Coinkite says you are **not** treated as at risk from this RNG issue alone if, when creating the seed, you entered **at least 50** fair, independent dice rolls through **Add Dice Rolls** (and those rolls stayed private). **99+** rolls push dice-only entropy near **256 bits**. Fewer than 50 rolls, or any uncertainty about how the seed was made, means migrate. A strong, unique **BIP-39 passphrase** (not the device PIN) adds an independent barrier. Short, quoted, or reused passphrases should not be treated as a fix. Coinkite still wants passphrase users to migrate to a fresh seed as soon as practical. Multisig setups and wallets that never used a Coldcard-generated seed sit outside the core warning, though investigators have focused on single-sig spends that look like predictable keys. ## What Coinkite wants you to do The preferred path is calm migration, not a panic send: 1. **Upgrade first** if you will generate a replacement seed on Mk4, Mk5, or Q: install **5.6.0** or **1.5.0Q** (or later) from the [official upgrade page](https://coldcard.com/docs/upgrade/), bless the firmware, then create a **new** seed. 2. **Verify** the new backup and the receive address on the device screen. Send a small test amount and confirm receipt before moving the rest. 3. **Keep the old backup** until the new wallet is proven. Rushing a migration is how people lose coins to typos and wrong-network mistakes. 4. **Mk3-only stopgap:** apply a long, random BIP-39 passphrase on-device, verify the new fingerprint, test-send, then migrate off Mk3 hardware when you can. Advanced users can build a dice-only seed on empty Mk3 **4.1.9** via **Import Existing → Dice Rolls** with **99+** rolls (that path does not use the device RNG). Bitcoin developer James O’Beirne, cited by The Block, put the urgent cohort in plain language: single keys generated on a Coldcard Mk3 between 2021 and 2023, without dice, passphrase, or multisig, should move funds soon, carefully. ## Takeaway Self-custody still means *you* own the failure modes. Hardware wallets remove exchange counterparty risk; they do not remove firmware and entropy risk. The practical checklist for LatAm and global holders who treat a Coldcard as their long-term vault is the same: confirm **which model and firmware** generated each seed, treat weak-entropy seeds as compromised for planning purposes, upgrade before minting a replacement key, and prefer multisig or passphrase + verified migration over a single rushed transfer. This is a security incident report, not investment advice. If your coins sit on an affected seed, follow Coinkite’s migration steps and verify every address on the device screen before you send. --- ## Keep reading - [How to plan gas so your stablecoins stay spendable](https://ptycoin.com/en/posts/2026-07-28-gas-planning-native-fees-stablecoins/index.md): USDT can land in a wallet and still be unspendable. You need the chain's own coin for fees. This guide shows how to keep TRX, ETH, SOL, and similar gas ready before you need it. - [How to pick the right network for USDT and USDC](https://ptycoin.com/en/posts/2026-07-21-pick-right-network-usdt-usdc/index.md): USDT and USDC exist on many chains. Wrong-network sends are a top permanent-loss risk. Use this checklist to match asset, network, fees, and wallet support before you transfer. - [How to receive crypto safely into a self-custody wallet](https://ptycoin.com/en/posts/2026-07-14-receive-crypto-safely-self-custody/index.md): Receiving is safer than sending, but the wrong network or a shared address still loses funds. This checklist shows how to generate, verify, and share a receive address correctly. --- Source: PTYcoin — https://ptycoin.com/en/posts/2026-07-31-coldcard-mk3-seed-entropy-warning-594-btc/. Free to read and cite with attribution to ptycoin.com. AI-usage terms: https://ptycoin.com/en/ai-usage/