BTC ETH Security

AI Could Crack Bitcoin and Ethereum's Security in Months, Not Years

2026-10-08

The crypto industry got a jolt this week. Justin Drake, an Ethereum researcher, raised an uncomfortable possibility: artificial intelligence might break the cryptography securing Bitcoin and Ethereum not in decades, but within months if things go very badly.

Here's the actual threat. Bitcoin wallets and Ethereum accounts rely on elliptic curve cryptography—two linked numbers, a secret private key and a public key. Computing a public key from a private key is trivial. Reversing it shouldn't be possible. Drake's concern is that AI could discover a mathematical shortcut that makes reversal practical on ordinary computers, years before quantum machines arrive.

OpenAI released 722 mathematical manuscripts from an unreleased AI model on October 6, tested against roughly 4,000 research problems. Drake flagged this on October 7. The speed of AI mathematical discovery had jumped roughly tenfold month-over-month. No attack exists yet. Drake wasn't calling for panic—he was calling for something he termed "bunker mode": a gradual, deliberate shift of holdings to fresh addresses whose public keys have never touched the blockchain.

Millions of Bitcoin addresses have already revealed their public keys through past transactions. On Ethereum, any account that's ever sent a transaction has exposed its key. Stablecoins, tokenized assets, the lot—all secured by the same maths. A practical break would theoretically expose a substantial slice of all cryptocurrency held in these addresses.

Vitalik Buterin weighed in quickly. He didn't dismiss Drake's warning. But he extended the problem further: lattice-based cryptography, the quantum-resistant alternative the industry has been developing and the U.S. standards body endorsed, could also weaken under AI-accelerated mathematical progress within two years. He compared it to how the general number field sieve forced RSA encryption to abandon smaller key sizes.

The recommended defense is operationally messy. Large holders should move funds to fresh addresses that have no public keys on-chain, depriving attackers of their starting point. Standard practice for paranoid holders, executed at network scale.

Buterin flagged something else, though. "I personally have lost more money in botched migrations than I have lost in all hacks combined," he wrote. A panicked, careless migration—moving funds badly during a crisis—could cost more immediately than AI ever would. His advice: reduce exposure where practical, but execute carefully.

Ethereum's roadmap is already shifting. The Ethereum Foundation formed a dedicated Post-Quantum Security team in January 2026 and has been running tests on hash-based validator signatures across multiple clients. The network targeted December 2029 for the switch to quantum-resistant cryptography. Drake's worst-case timeline—years earlier—has forced developers to reconsider.

Hash-based signatures, which turn data into unforgeable digital fingerprints, look like Ethereum's best long-term answer. They leave fewer obvious mathematical openings for shortcuts. Buterin acknowledged even hash systems could theoretically be attacked, but the design seems more resilient.

The reaction has split. Some researchers argued Drake's timeline isn't credible or the risk isn't urgent. Others noted that Solana and newer blockchains keep public keys hidden by design, sidestepping the exposure. What nobody disputes: AI is genuinely accelerating mathematical discovery in ways that could hollow out assumptions baked into cryptographic systems designed years ago.

Source & further reading:

Sources