Coldcard Hack: Nearly $114 Million in Bitcoin Threatened by AI-Powered Attacks

Bitcoin self-custody, long considered the ultimate shield against exchange hacks, is now firmly in the crosshairs of attackers. A major incident involving Coldcard wallets is sending shockwaves through the crypto community.

Potential losses are now approaching $114 million — and artificial intelligence is believed to be at the core of the attack mechanism. This is a warning signal that is impossible to ignore.

Here is a full breakdown of a threat that is redefining security standards for Bitcoin self-custody holders.

Coldcard Under Fire: How Did the Attack Work?

Coldcard is one of the most respected hardware wallets in the Bitcoin ecosystem, favored by advanced users for its robustness and air-gapped architecture. Yet the ongoing incident demonstrates that no device is entirely immune to sophisticated attacks.

Based on the information available so far, the attack does not directly target the wallet’s firmware. Instead, it exploits peripheral vectors — potentially linked to supply chain attacks, compromised seed phrases, or third-party management interfaces. The involvement of AI in this attack vector marks a clear turning point: generative tools now make it possible to craft hyper-personalized phishing campaigns, simulate legitimate interfaces with unsettling precision, and automate large-scale attacks against specific targets.

Exposed funds are estimated at close to $114 million in Bitcoin, according to data aggregated by on-chain security researchers. This figure places the incident among the most significant non-custodial wallet compromises of 2025.

AI — The New Offensive Weapon Against Crypto: What It Really Changes

The use of artificial intelligence in crypto cyberattacks is no longer a theoretical scenario. Recent reports from firms such as Chainalysis and CertiK document a sharp rise in AI-powered tools within hackers’ arsenals: fake exchange websites, audio and video deepfakes impersonating legitimate projects, and now precise targeting of hardware wallet holders.

In the Coldcard case, AI is believed to have industrialized social engineering — automatically identifying high-net-worth crypto profiles across social media, specialized forums, and leaked databases, then delivering perfectly calibrated fraudulent communications to those targets. This level of personalization was previously reserved for state-sponsored APT (Advanced Persistent Threat) attacks.

For Bitcoin self-custody holders, the implications are immediate and direct:

• Never enter your seed phrase on a connected device, regardless of who is asking
• Always verify the physical integrity of your hardware wallet upon receipt
• Be suspicious of any communication — even one that appears official — requesting urgent action on your wallet
• Enable the BIP39 passphrase as an additional layer of security

What Impact Will This Have on Trust in Bitcoin Self-Custody?

This incident comes at a time when self-custody is experiencing a massive resurgence in interest, driven by post-FTX distrust toward centralized exchanges. The mantra “not your keys, not your coins” has convinced millions of users to move their funds into non-custodial solutions. But the growing sophistication of attacks raises a fundamental question: is self-custody truly accessible to all types of investors?

Security experts are unanimous on one point: the hardware wallet remains the best protection available today — provided strict protocols are followed. The real weak link is not the device itself, but the human and digital environment surrounding it. AI amplifies this risk by making social attacks nearly undetectable for the average user.

Coinkite, the company behind Coldcard, has yet to publish an official statement detailing the exploited vectors or the corrective measures being considered. The Bitcoin community is waiting for answers — and guarantees — regarding the security of one of its most trusted tools.