Concerns arise as Ethereum researcher indicates that AI could compromise the security of Bitcoin and Ethereum wallets much sooner than anticipated.
By Shaurya Malwa|Edited by Stephen AlpherOct 8, 2026, 7:47 a.m. EDT4 min readMake preferred on ShareShare this articleCopy linkX (Twitter)LinkedInFacebookEmailMake preferred on Ethereum researcher warns of AI threats to Bitcoin and Ether. (Chris Ried/Unsplash)SummaryShow- Justin Drake, an Ethereum researcher, has urged the crypto community to adopt “bunker mode” in response to potential AI threats that could undermine the security of Bitcoin and Ethereum wallets.
- While no actual attacks have been reported, Drake suggests that significant holders should gradually transfer their assets to new addresses that have not been exposed on-chain.
- Vitalik Buterin, co-founder of Ethereum, echoed these concerns and noted that AI could also jeopardize some quantum-resistant cryptography.
Justin Drake, a researcher in the Ethereum community, has advised the crypto sector to enter a "bunker mode" due to the looming threat that artificial intelligence might breach the mathematical safeguards protecting Bitcoin and Ethereum wallets in a worst-case scenario "in months, not years."
In a post on X, Drake called on the blockchain industry to start planning for this potential risk, suggesting that major holders begin moving their funds to new addresses gradually.
Wallets utilize a private key, a secret number that authorizes payments, paired with a public key that the network uses to validate these transactions. Although producing a public key is straightforward, deriving the private key from it is designed to require an impractical level of computational effort.
Drake's concern is that AI might discover a method to facilitate this process, enabling an attacker to retrieve the secret key and access the funds without the advanced quantum computers that are currently being anticipated.
This development could render a significant portion of cryptocurrency vulnerable. Numerous Bitcoin assets are stored in addresses with visible public keys, as previously reported by CoinDesk. Likewise, on Ethereum, any account that has executed a transaction has exposed its key, which governs stablecoins and other tokenized resources on the network as well.
No verified attacks against Bitcoin or Ethereum wallet keys have been documented thus far, nor have any relevant research findings surfaced in CoinDesk's investigations.
AI as a Crypto Risk
This alert comes in the wake of OpenAI's recent release of 722 mathematical manuscripts generated by a model tested on around 4,000 research problems. Some of these results have computer-verifiable proofs, while others may contain errors, according to the company.
The manuscripts were produced by a model that OpenAI claimed had solved the Navier–Stokes problem, one of the seven Millennium Prize challenges in mathematics. Each result required an average computational effort equivalent to about three hours of ChatGPT Pro reasoning.
Within a day, an external researcher replicated the computer verification on one result, a new limit on the speed of multiplying large grids of numbers, a problem that has been explored by mathematicians since 1969, and confirmed its validity.
Drake explained that the mathematical principles underlying Bitcoin and Ethereum wallet signatures, which are based on elliptic curves, have predictable patterns that a sufficiently powerful AI could exploit. Hash functions, designed to convert any data into a fixed-length digital signature that cannot be reversed, aim to obscure information with minimal patterns.
AI-facilitated attacks on cryptocurrency systems have already resulted in financial losses.
In December, researchers from Anthropic demonstrated that advanced models could create working exploits against simulated DeFi contracts. In late July, a group known as the Bitcoin Red Team employed AI models to examine 390 Bitcoin software projects in roughly 27 hours, identifying nearly 5,000 potential vulnerabilities, 85 of which were categorized as critical.
On July 30, an attacker exploited a five-year-old firmware flaw in Coldcard hardware wallets, resulting in the theft of at least 1,367 BTC. The manufacturer, Coinkite, suggested that AI may have aided in discovering this vulnerability.
Shortly after, BTCPay Server confirmed that attackers had pilfered funds from merchants' Lightning nodes through a vulnerability first revealed in an AI-assisted audit, and on August 27, Core Lightning developers issued an emergency alert after AI-generated bug reports uncovered genuine weaknesses in their software.
Researchers have also employed AI coding agents to enhance calculations relevant to future quantum attacks, although such efforts still necessitate quantum hardware and only address part of the attack strategy, as reported by CoinDesk in September.
Read More: Crypto researchers cut Bitcoin and Ethereum quantum attack estimate by 50%
The timelines for these developments do not align. The Ethereum Foundation is aiming to transition the network to quantum-resistant cryptography by December 2029.
However, Drake's worst-case scenario suggests a classical breach could occur years prior.
Limits of Quantum Resistance
Vitalik Buterin, co-founder of Ethereum, acknowledged the genuine risk and expanded it to include some of the alternatives that developers are relying on.
Some quantum-resistant designs utilize lattice-based cryptography, which is believed to be difficult for both classical and quantum computers, and this approach forms the basis of a digital-signature standard endorsed by the U.S. National Institute of Standards and Technology.
"There is a good chance that the concrete security of lattices will take serious hits from the next two years of AI math," Buterin remarked on X. “If AI can provide us with 50 years of mathematical advancement in just two years, then it is plausible that this could include significant improvements in our capability to break lattice-based systems."
Ethereum's long-term rebuilding plans increasingly favor hash-based signatures, which convert data into digital fingerprints designed to be challenging to reverse. Buterin perceives fewer opportunities for unexpected vulnerabilities in these structures, though he admitted they could also be susceptible to attacks.
Drake advised that sophisticated holders should take the first step, moving their funds to addresses whose public keys have never been publicly displayed on-chain, thereby concealing the essential starting point for any attack. Conversely, Buterin supported minimizing public-key exposure where feasible but cautioned against hasty migrations, as errors during such transitions could lead to substantial losses for users. "I personally have lost more money in botched migrations than I have lost in all hacks combined," he stated.