以太坊的贾斯汀·德雷克警告称，人工智能可能在几个月内破解加密货币钱包

Disclaimer: Content does not constitute investment advice. Trading involves risks—please invest with caution!

Ethereum researcher Justin Drake: AI may crack crypto wallets within months, and the industry needs to launch a "bunker model"

Ethereum researcher Justin Drake urged the cryptocurrency industry to be prepared for scenarios where artificial intelligence may break through network asset protection signature algorithms such as Bitcoin and Ethereum within months. This risk is far ahead of the expected arrival of quantum computers.

AI threats exacerbate the urgency of cryptographic security

Drake warns that recent advances in artificial intelligence may soon allow attackers to crack the private key behind wallets holding Bitcoin, Ethereum and other tokens. He recommends that large currency holders start moving assets to addresses where public keys have never been exposed on the chain to mitigate potential risks.

No actual attack cases have been confirmed. However, Drake emphasized proactive measures to protect huge amounts of money and called on the industry to calmly prepare for the so-called "bunker model."

Drake wrote on Platform X (formerly Twitter):"Today I call on the blockchain industry to calmly start planning a 'bunker model'."

The mathematical basis for protecting cryptocurrency wallets relies on elliptic curve cryptography, which was designed so that deriving private keys from public keys requires infeasible computing resources. Drake warns that powerful artificial intelligence models may soon provide a shortcut that allows attackers to reveal private keys using conventional hardware.

Currently, millions of bitcoins are stored at addresses where public keys have been exposed. On the Ethereum network, any account that has conducted a transaction reveals its public key, which could increase vulnerability if AI-assisted attacks become feasible in practice.

OpenAI's development and recent attacks

Drake's warning follows OpenAI's release of 722 mathematical manuscripts generated from undisclosed AI models. The documents include solutions to complex problems such as the Navier-Stokes equations, signaling AI's growing ability to solve highly specialized mathematical challenges. Some results have been quickly verified by external researchers.

Recent accidents show that AI-based systems are already able to identify vulnerabilities in smart contract code and wallet software. For example, Anthropic's research shows that cutting-edge models can generate viable utilization code for DeFi contracts. In July this year, the Bitcoin Red Team used AI to scan nearly 400 Bitcoin software projects and found approximately 5,000 possible errors, of which 85 were considered critical.

In a high-profile case, Coldcard's hardware wallet was compromised due to a firmware vulnerability, resulting in the theft of at least 1,367 bitcoins. Manufacturer Coinkite said AI-assisted analysis may have discovered the flaw. In addition, attacks on BTCPay Server and Core Lightning infrastructure are also related to vulnerabilities identified in vulnerability reports generated through AI.

Researchers are also using AI-encoded agents to improve calculations related to quantum attacks on cryptographic systems, although these efforts still require quantum hardware support and are still in the theoretical stage in nature.