Cryptographer Warns AI Could Crack Crypto Wallet Math

Artificial intelligence might pose an unforeseen challenge to the mathematics that safeguards Bitcoin and Ethereum. Researchers are asking if AI could find a shortcut enabling hackers to crack encryption and digital signatures that secure the internet and cryptocurrencies.

This worry comes after OpenAI's October 6 announcement that an unreleased AI model generated 722 math manuscripts across 372 result groups.

The AI went after about 4,000 unsolved problems, and the results are still under review.

This development prompted a warning from Johns Hopkins cryptographer Matthew Green.

I think we might lose public key cryptography.

— Matthew Green (@matthew_d_green) October 8, 2026

Green later stated he was not aware of any imminent cryptographic breakthrough. Rather, he fears AI might uncover mathematical weaknesses that human researchers overlooked for decades.

What Risk Does AI Pose to Bitcoin and Ethereum Wallets?

Both Bitcoin and Ethereum use elliptic-curve cryptography to secure transactions. The underlying math makes it nearly impossible to derive a private key from a public one.

Should AI find a much quicker route to solve that problem, attackers might recover private keys and drain funds.

Wallets with publicly exposed keys would be especially at risk, as would Ethereum accounts that have previously signed transactions.

The fallout could reach online banking and encrypted internet communication as well.

“What I would argue is that most of the open problems solved by OpenAI with their new models were actually rather close to a solution, conceptually speaking, and that mathematicians working in the field would have come to solutions soon with new tools in their hands.“

Today's cryptography typically aims for very high security margins. Green proposed that AI findings might show systems thought to have 128-bit security actually only afford 96 or 108 bits.

A 20-bit drop would cut the cost of an attack by about a factor of a million. Yet even with that, an attack might stay too expensive to execute.

Larger keys could mitigate the issue, but Green asked how experts would decide on the correct size.

“But crank them up to what?”

Post-Quantum Cryptography May Also Face Scrutiny

The worry extends to systems designed as a defense against quantum computers.

In 2024, the U.S. National Institute of Standards and Technology (NIST) gave the green light to new post-quantum standards, among them ML-KEM and ML-DSA.

These standards depend on lattice-based math. If AI discovers significantly better attacks on those problems, some planned protections might need to be adjusted.

But Blockstream CEO Adam Back dismissed the warning as "dubious."

Blockstream's researchers have created SHRINCS, a hash-based signature scheme that uses SHA-256. Its signatures are as small as 324 bytes, and the system has been shown to work on Bitcoin's Liquid sidechain.

Bitcoin would still need a protocol upgrade to implement SHRINCS. Moreover, hash-based signatures cannot directly substitute for public-key encryption and key exchange.

A number of investors have begun taking precautions.

Charles Edwards, founder of Capriole Investments and a member of BeInCrypto’s Market Intelligence Expert Council, stated he has quit using DeFi and trimmed his wallet and account count.

If AI compromises mainstream encryption in any way, there will be chaos.

Currently, no publicly proven AI attack can crack Bitcoin or Ethereum's core cryptography.

The issue is whether the industry can update its security fast enough if AI eventually finds such an attack.