Lopp Says AI Crypto Fears Overshadow Real Software Risks
Jameson Lopp, co-founder of Casa, rejected claims that artificial intelligence is close to breaking blockchain encryption. He argued that the crypto sector should prioritize existing software vulnerabilities, which attackers can exploit more easily. The debate followed remarks by Justin Drake and Haseeb Qureshi that new AI models might crack ECDSA signatures sooner than expected.
Jameson Lopp, Casa’s co-founder, pushed back against claims that AI is nearing a breakthrough against blockchain encryption. The discussion grew after Ethereum Foundation’s Justin Drake and Dragonfly’s Haseeb Qureshi said Anthropic and OpenAI models might crack ECDSA signatures within months. Drake pointed to Claude Mythos Preview finding a flaw in post-quantum HAWK in 60 hours.
Lopp instead highlights practical weaknesses: AI can scan wallet and app code for human mistakes. Google Threat Intelligence data indicates the monthly count of disclosed software flaws nearly doubled over the past year, while Coldcard seed-generation issues could be found in minutes. Vitalik Buterin sees long-term crypto risk but warns against overreaction. No active ECDSA key crack has been confirmed.
The debate may affect crypto users, developers, and custody providers most directly. If AI-assisted tools make software bugs easier to find, wallet and exchange security could improve through faster audits, but poorly maintained projects may face more exploits. If cryptographic fears are overstated, resources might be diverted from practical fixes. Public confidence in digital assets could depend less on hypothetical AI breakthroughs than on whether everyday vulnerabilities are addressed.