If you thought keeping Bitcoin on a hardware wallet made you untouchable, the Coldcard hack should shake that confidence. Potential losses have now crept past $114 million, and the attack vector is chillingly modern: AI-generated phishing tailored to each victim’s identity. This isn’t a random exploit—it’s a systematic dismantling of the ‘cold storage equals safety’ dogma.
The attack doesn’t break Coldcard’s encryption. Instead, it uses AI-driven social engineering to trick users into revealing their seed phrases or signing malicious transactions. Think of it as a bespoke con job, where the scammer has access to your browsing history, wallet addresses, and even your social media activity. The AI crafts a message that looks exactly like a legitimate Coldcard firmware update or a security alert. One click, and your funds are gone.
According to on-chain data from chainalysis and public reports, the largest single loss so far is $85.3 million in Bitcoin, traced to a single address that drained a cold wallet in late February. The remaining ~$28.7 million is spread across dozens of smaller victims, with the total climbing weekly. And the really scary part? Most victims didn’t even know their seed phrase had been compromised until days later.
How the Coldcard Hack Works
The attack relies on a two-step process. First, the attackers scrape data from public sources—exchange profiles, forum posts, even LinkedIn—to build a detailed profile of the target. Then, they use a large language model (LLM) to generate a hyper-personalized phishing email or direct message. The message might reference your specific Coldcard model, your previous transactions, or even the date you bought the device. It asks you to ‘update firmware’ or ‘verify your seed backup’ by entering your 24 words into a fake website that looks identical to the real Coldcard portal.
Once the seed is captured, the attacker can sweep all funds from that wallet. Coldcard’s hardware itself is not compromised—it’s the human element that’s exploited. And because AI can now mimic tone, grammar, and urgency perfectly, even experienced users are falling for it.
This is a massive shift. In 2022, the Ledger data breach exposed customer names and addresses, leading to phishing attacks. But those were blunderbuss—sent to thousands in hope a few bite. The Coldcard attack is a sniper rifle: tailored, patient, and devastatingly effective.
AI Is the Silent Force Behind the Surge
Look, I’ve tracked crypto long enough to know that every bull run brings new attack vectors. But AI changes the game entirely. It scales social engineering to industrial levels without requiring human hackers to write each message. As ZeroStack warned after its own $82.5 million loss, AI-powered attacks are now a survival risk for anyone in self-custody. The Coldcard hack is just the proof.
What’s worse, the attackers are using open-source LLMs to generate these messages. They don’t need a custom botnet or zero-day exploits. They just need a victim list and a few hours of compute time. The cost per attack? Pennies.
And it’s not just Coldcard. Trezor, Ledger, and even paper wallet users are seeing similar patterns. The difference is that Coldcard’s marketing has long emphasized ‘maximum security’—making its users prime targets. The irony stings.
What This Means for Your Cold Storage
If you hold a Coldcard right now, here’s the hard truth: your device is still secure, but your process might not be. The attack doesn’t steal from the hardware; it steals from your decision-making. So what can you do?
First, never enter your seed phrase into any website or app. Coldcard, like all reputable hardware wallets, will never ask for your seed online. If a message directs you to a site that asks for it, that site is malicious—AI-generated or not.
Second, use a passphrase (BIP39). Even if attackers get your 24-word seed, a passphrase adds an extra layer. Without it, they can’t access the funds. This is a simple, effective countermeasure.
Third, verify all firmware updates manually. Download from the official Coldcard website only, and check the SHA256 checksum. Don’t trust links in emails or DMs—ever.
Fourth, consider multi-signature setups. If you have significant holdings, spread the risk across multiple devices and signers. That way, a single compromised seed won’t drain everything.
For institutional holders, the stakes are even higher. As we noted when Bitcoin dropped below $63K, market volatility often coincides with spikes in theft attempts. The Coldcard hack is compounding that trend.
The Bigger Picture: Self-Custody vs. Centralization
Every time a self-custody hack makes headlines, the pendulum swings back toward centralized exchanges. ‘Why bother with cold storage if it can be hacked?’ people ask. And sure, Coinbase or Binance have insurance and fraud teams—but they also have single points of failure, regulatory risks, and their own history of hacks.
The real lesson here isn’t that self-custody is dead. It’s that self-custody requires constant vigilance. The threat model has evolved. AI means you can no longer rely on ‘just don’t click weird links.’ You need to assume every communication is a potential attack, even if it looks flawless.
What’s the likely endgame? Either hardware wallet makers integrate AI-level defense (like automated verification of messages), or users will gravitate toward custodial solutions with AI fraud detection. The market is already moving—Robinhood’s UK expansion shows that regulated, user-friendly platforms are gaining ground. But that comes with trade-offs in sovereignty.
My read: the Coldcard hack is a watershed moment. It proves that AI can crack the hardest nut in crypto—the unconnected, offline wallet. The only defense is a paranoid workflow: never trust the message, verify everything, and assume your seed is one mistake away from exposure.
As the losses climb toward $114 million, the industry is waking up. But for the victims, it’s already too late. The next wave of AI-driven attacks won’t target Coldcard alone. They’ll target you. And the only question is whether you’ve hardened your process enough.
Frequently Asked Questions
Is Coldcard still safe to use?
Yes, the hardware itself remains secure. The hack exploits user behavior, not the device’s encryption. If you never enter your seed phrase online and verify all firmware updates manually, your funds are safe.
How can I protect against AI-based phishing attacks?
Use a passphrase (BIP39) with your seed, never click links in unsolicited messages, and always verify the official domain. Consider multi-signature wallets for large amounts.
What are the actual losses from the Coldcard hack so far?
As of late March 2025, on-chain data tracks approximately $85.3 million from a single major theft and another $28.7 million from smaller incidents, totaling around $114 million. The number is expected to rise.
