Matthew Green warns AI could undermine confidence in public-key encryption

Cryptographer Matthew Green has warned that AI could erode confidence in public-key encryption algorithms, which secure much of modern communication.

He puts the chance that such algorithms lose confidence at 15%, and a 1% chance that cryptography is impossible in the Minicrypt world, a theoretical model in which one-way functions exist but public-key cryptography does not. Green, as reported by Simon Willison, argues that AI can produce unexpected breakthroughs far faster than humans can replace cryptographic standards. He therefore says that recovering from such a failure would require preparing replacement systems in advance.