Could AI Break Cryptography Soon? Yes, but 'Bunker Mode' May Not Be the Solution

The episode centers on whether AI could break blockchain cryptography sooner than quantum computers, and whether that should push the industry toward “bunker mode” and hash-based systems. The host frames the debate around Justin Drake’s warning, Alex Pruden’s and Conor Deegan’s view from Project 11, and the pushback from cryptographers who say there is no evidence elliptic-curve schemes are failing today.

Alex argues that AI and quantum are separate but overlapping threats, and that public-key cryptography rests on hard math assumptions that may not hold forever. Conor says elliptic-curve signatures remain safe today, but the industry should not wait for a proof of failure before planning migrations. Both argue that cryptography should become more “agile,” with protocols ready to move between signatures, lattices, and hash-based schemes rather than assuming one standard will last indefinitely.

They also distinguish between signatures and encryption. Hash-based signatures look attractive because they avoid much of the algebraic structure AI might exploit, but there is no clean hash-based replacement for all uses of public-key encryption. Alex says that means privacy systems such as Zcash are more exposed in some scenarios, while Conor emphasizes that blockchain assets are uniquely vulnerable because a private-key break can directly move funds with no intermediary to stop it.