Artificial intelligence may end up pressuring more than software development and content generation. It could also challenge the cryptography that underpins the security of the crypto industry.
Ethereum co-founder Vitalik Buterin said the sector should take seriously the risks tied to AI-accelerated mathematical research. If AI helps researchers discover new algorithms and attack methods, the practical security lifespan of existing cryptographic systems could turn out to be shorter than previously assumed.
At the same time, Buterin said there is no evidence that AI has broken the wallet signatures used by Bitcoin or Ethereum, so there is no reason for users to rush into panic-driven asset transfers.
Buterin says the risk is real, but not a reason to scramble
Buterin made the remarks while responding to warnings from Ethereum Foundation researcher Justin Drake on cryptographic security.
Drake argued that if AI sharply improves the pace of mathematical research, the next major threat to current cryptography may arrive before large-scale quantum computers do.
Buterin agreed that the possibility deserves serious preparation, but said users should not feel compelled to move all funds into new wallets right now.
If moving assets is easy, placing part of one’s holdings in fresh addresses that have never sent a transaction can be an extra safety measure. Still, he warned that wallet migration carries its own risks. A wrong address, a failed private-key backup, or signing the wrong transaction can all lead to permanent losses.
The concern is not AI directly guessing private keys
The issue is not that ChatGPT or another large language model can directly compute a user’s private key. The real concern is that AI may speed up mathematical research and algorithm discovery.
Bitcoin and Ethereum both rely heavily on elliptic-curve cryptography. These systems are considered secure because, with currently known algorithms, deriving a private key from public information is computationally infeasible within a practical timeframe.
If AI helps researchers find previously unknown mathematical shortcuts that reduce the work needed to solve those problems, the strength of existing cryptographic systems could weaken. For now, though, there is no public evidence that such a breakthrough has happened.
Ethereum’s official security materials also say current quantum computers are still not capable of breaking the main cryptography used by Ethereum, meaning ordinary users do not need to take emergency action at this stage.
Could AI arrive before quantum computing as a threat?
When crypto’s cryptographic risks come up, the discussion usually centers on quantum computing. In theory, a large fault-tolerant quantum computer could use Shor’s algorithm to break the elliptic curve digital signature algorithm, or ECDSA, that is widely used today.
The problem is timing. No one knows when a quantum computer with that level of capability will actually exist.
Drake raised another possibility: before quantum hardware reaches that point, AI may already have helped uncover new mathematical methods that lower the cost of breaking current systems. He went further and said the industry should not completely rule out the possibility that such a breakthrough could emerge in months rather than years.
That remains a risk scenario, not a confirmed timetable. There is currently no public research showing that ECDSA is close to being breakable at scale with existing GPUs.
Even post-quantum cryptography may not be beyond review
One of the more notable parts of Buterin’s comments is that his concern is not limited to elliptic-curve cryptography.
He also said lattice-based cryptography, long viewed as a major post-quantum option, could face renewed security scrutiny if AI materially speeds up mathematical discovery.
Lattice-based cryptography is one of the main research tracks in post-quantum cryptography worldwide. The U.S. National Institute of Standards and Technology, or NIST, has already standardized some related algorithms. The goal is to keep cryptographic systems secure even in a future with mature quantum computers.
But Buterin’s point is that every cryptographic system ultimately rests on the assumption that certain mathematical problems remain hard. If AI greatly increases the speed at which humans discover new algorithms, some assumptions now treated as relatively safe may need to be tested again.
For the crypto industry, the long-term question may not be limited to quantum resistance. It may also be about building cryptographic systems that are more resilient to different kinds of mathematical breakthroughs.
Why unused addresses may offer an extra layer of protection
Both Buterin and Drake referred to fresh addresses that have never been used, and the reason comes down to whether the public key has been exposed.
In a standard Ethereum account, the address is derived by hashing the public key. If an address has never sent a transaction, the chain usually shows only the address itself, not the full public key.
Once that address sends a transaction, signature data appears on-chain and the public key can be recovered from the signature. Under a hypothetical future in which a new technique can efficiently derive private keys from public keys, an address that has never sent a transaction would still benefit from an extra layer of hash-based protection.
That is why some cryptography researchers suggest that users with large holdings may consider storing long-term cold funds in fresh addresses that have never sent a transaction. Even so, this is a defensive measure, not proof that current wallets are already unsafe.
Ethereum is studying a more conservative cryptographic path
Buterin has spent recent years pushing Ethereum toward cryptographic designs aimed at stronger long-term security. One direction is greater use of hash-based constructions.
Compared with systems that depend on elliptic curves or lattice structures, hash-based signatures usually rest on simpler security assumptions that have been studied for a long time.
Ethereum is studying hash-based signature approaches including leanXMSS and SPHINCS as part of its future post-quantum security architecture.
The Ethereum Foundation also set up a dedicated post-quantum security research team in 2026 to gradually reduce Ethereum’s reliance on cryptographic techniques that could be affected by quantum computing.
Buterin’s latest comments on AI add another reason for that work. Cryptographic upgrades, in his view, should account not only for quantum computers, but also for the possibility that AI could speed up mathematical breakthroughs.