Solana Rejects Crypto “Bunker Mode” as AI Security Debate Intensifies
Solana developers are challenging calls for cryptocurrency users to move their funds into new wallets after Ethereum Foundation researcher Justin Drake warned that advances in artificial intelligence could accelerate attacks on cryptographic systems. The discussion has divided researchers over whether the industry faces an urgent security threat or should focus on long-term preparation.
Drake’s October 7 warning urged crypto holders to consider moving assets to fresh addresses that have never signed transactions, keeping their public keys hidden behind cryptographic hashes. He argued that elliptic-curve digital signatures could potentially become vulnerable within “months, not years” in a worst-case scenario, even before quantum computers become capable of breaking them.
Solana Foundation technology executive Jacob Creech disputed the need for this approach, arguing that Solana’s account design offers a potential path to migrate users to new cryptography. Meanwhile, Coinbase cryptographer Yehuda Lindell questioned the evidence behind Drake’s warning, while Ethereum co-founder Vitalik Buterin said the underlying risks deserve serious investigation.
Why Researchers Are Discussing “Bunker Mode”
Drake’s proposal focuses on the Elliptic Curve Digital Signature Algorithm (ECDSA), a cryptographic system used to authorize transactions on Bitcoin and Ethereum. Digital signatures allow networks to verify that a transaction was authorized by the holder of a private key without revealing that key.
In some address formats, spending funds reveals the associated public key. If an attacker could derive the private key from an exposed public key, the attacker might be able to authorize unauthorized transactions. Moving funds to unused addresses can reduce certain forms of exposure, although it cannot eliminate every security risk.
Drake’s concern is that AI systems could help researchers discover mathematical shortcuts that conventional analysis has missed. He pointed to developments involving integer multiplication, the 3SUM problem and the Erdős unit distance conjecture as reasons to reassess assumptions about how difficult some mathematical problems are.
His recommendation has triggered debate over whether users should take protective action now or wait for stronger evidence of a practical cryptographic attack.
Vitalik Buterin Urges Caution and Further Research
Ethereum co-founder Vitalik Buterin rejected the idea that users should immediately rush to migrate their funds, but agreed that AI-driven mathematical advances could create new challenges for cryptography.
In his post on X, Buterin argued that researchers should consider potential AI-related weaknesses alongside the better-known risks associated with quantum computing. He also warned against assuming that cryptographic systems designed to resist quantum attacks will necessarily remain secure against future mathematical breakthroughs.
Buterin specifically highlighted lattice-based cryptography, including ML-DSA and technologies used in fully homomorphic encryption. These approaches are relevant to post-quantum security, but their inclusion in the discussion does not mean they have been broken. The concern is whether improved mathematical techniques could eventually weaken existing security estimates.
He also cautioned that poorly executed wallet migrations can create immediate problems. Buterin said he had personally lost more money through botched migrations than through hacks combined, underscoring the need to balance precaution with operational safety.
Coinbase Cryptographer Disputes the Evidence
Yehuda Lindell, head of cryptography at Coinbase, described Drake’s warning as “the very definition of FUD,” arguing that no publicly demonstrated evidence shows that AI has broken the mathematical assumptions underlying elliptic-curve cryptography.
Lindell’s position is that proving difficult theorems with AI does not automatically mean the technology can solve problems such as the elliptic-curve discrete logarithm problem. He also disputed the claim that elliptic curves are necessarily more vulnerable to AI-assisted attacks than cryptographic hash functions.
His concerns extend beyond cryptocurrency. A practical break of widely used public-key cryptography could affect encrypted web connections, digital certificates, software signing and other systems that depend on digital signatures. The potential consequences would therefore reach far beyond individual blockchain networks.
Other researchers disagree about how much reassurance the absence of a public attack should provide. Johns Hopkins cryptography researcher Matthew Green has warned that AI could improve cryptanalytic techniques in ways that reduce the effective security of widely deployed systems. These remain concerns about potential future developments, not proof that major crypto signature schemes have already failed.
Why Solana Says It Has a Different Migration Path
Jacob Creech, vice president of technology at the Solana Foundation, argued that Solana users do not necessarily need to follow Drake’s proposed “bunker mode” approach. Solana uses Ed25519 signatures, with signing keys derived from a secret seed through a hashing process.
Creech said the seed itself does not appear onchain, which could provide a route for legitimate account owners to prove knowledge of their secret and migrate to a different signature scheme if the existing curve were compromised. Such a process would still depend on the design and implementation of an appropriate upgrade.
An experimental proof of concept is available in the Anza cryptography repository. Its existence demonstrates work on a possible migration mechanism, but does not establish that a complete network-wide recovery plan has been tested or deployed.
The broader distinction is that blockchain networks have different account structures, signing mechanisms and upgrade processes. A wallet-security recommendation that is useful for one chain may not be directly applicable to another.
AI Risk Could Extend Beyond Crypto
The debate also raises questions about how much of the internet relies on mathematical assumptions that future AI systems could challenge. Public-key cryptography is used in financial services, encrypted communications, software distribution and digital identity systems, making a major breakthrough potentially consequential across many industries.
Solana co-founder Anatoly Yakovenko has argued that AI would not simply overturn mathematical problems such as discrete logarithms or prime factorization by default. A more targeted possibility is that AI could uncover weaknesses in specific cryptographic parameters, implementations or groups without invalidating the underlying mathematical foundations entirely.
That distinction matters because discovering an implementation weakness is different from developing a general method for breaking an entire class of cryptographic systems. The appropriate response would depend on the nature, reliability and practical impact of any demonstrated attack.
What Crypto Users Should Take Away
For now, there is no publicly demonstrated evidence in the material discussed here that AI has broken the elliptic-curve cryptography protecting Bitcoin or Ethereum. Researchers disagree about how much weight to place on that absence of evidence, but the debate highlights the value of developing migration plans before a crisis emerges.
Crypto holders should be cautious about rushing into transfers based solely on an unverified threat. Moving assets can introduce risks of its own, including mistakes, malicious destination addresses and compromised recovery information. Users should verify wallet instructions and understand the security model of their particular blockchain before changing how they store funds.
The central issue is preparation: AI may change the pace of cryptographic research, but the industry still needs reproducible technical evidence to distinguish a credible attack from a theoretical possibility. Solana’s proposed migration path and Ethereum’s ongoing post-quantum work represent different approaches to the same challenge—maintaining security as cryptographic assumptions evolve.
