The theoretical threat of artificial intelligence hackers attacking blockchain encryption has been seriously exaggerated, and there is no need for the encryption industry to panic. Casa, a well-known Bitcoin developer and co-founder of Jameson Lopp, has come to this conclusion and urges the community to calmly assess the true nature of the AI threat.
The origin of this major discussion stemmed from a statement made by Justin Drake, a researcher at the Ethereum Foundation; Haseeb Qureshi of Dragonfly also expressed support. According to these two individuals, the new models proposed by Anthropic and OpenAI might be able to crack the fundamental ECDSA signature algorithm within “a few months rather than several years.”
Drake citing Claude Mythos Preview's neural network claims that the model discovered a vulnerability in the post-quantum HAWK algorithm within 60 hours, therefore he urges investors to urgently prepare for " bunker mode ".
However, Jameson Lopp, who has been dealing with the consequences of AI attacks on Bitcoin infrastructure over the past few months, believes that concerns about fundamental cryptography mathematics being soon broken have been exaggerated and are not very relevant to reality. What the theoretical community fears is whether the foundations of encryption will be hypothetically breached, while hackers are actually using AI for more practical purposes—automatically searching for common human-made vulnerabilities in application code and wallet firmware.
Lopp emphasized when commenting on broader concerns: 'Worrying about cryptography being cracked is a bit premature – we have more urgent practical issues to deal with.' In his view, the industry should first resolve the current operational problems, and 'theoretical future issues can be discussed at a later time.'
Why is “bunker mode” considered too early?
The views of Lopp are supported by the data from Google Threat Intelligence: In the past year, the monthly number of disclosed software vulnerabilities has almost doubled. Neural networks have become the ideal scanners, capable of detecting implementation-level flaws in just a few minutes – such as the weakness in the seed phrase generation of the Coldcard wallet, which would have otherwise taken humans weeks to discover.
Lopp believes that, at a time when AI is easily exposing poor-quality software, focusing attention on cracking ECDSA is counterproductive.
Ethereum co-founder Vitalik Buterin also joined this discussion. Although he acknowledges that AI poses a long-term risk to cryptography, he also believes that taking action too early may be more dangerous than the threat itself. Buterin joked that he has lost more money due to wallet updates going wrong than from all hacker attacks combined.
Rather than "retreating to a bunker," it's better to take basic safety and hygiene measures: collect multi-signature signatures off-chain, and avoid sending transactions from savings addresses in order to hide the public keys of those addresses.
As of now, there are no records indicating that the active ECDSA key has ever been successfully cracked, which further confirms that from a password analysis perspective, the threat posed by AI to Bitcoin is currently being seriously exaggerated.
In the view of Jameson Lopp, the winners of this competition will not be those who create obstacles for the hypothetical future "super AI", but rather those who start to address existing software security issues right now.
