A16z Crypto said AI has not yet broken modern cryptography, even as it may help accelerate security reviews. According to ChainCatcher, the firm invited Stanford University professor Dan Boneh, a16z crypto partner Eddy Lazzarin, and other experts to discuss AI’s threat to cryptography.

The experts said recent AI progress in mathematics has not produced cryptographic breakthroughs and has not shown that any specific cryptographic assumption has become less secure. Modern cryptography relies on the assumption that certain mathematical problems do not have efficient algorithms, and if that assumption holds, even powerful AI cannot discover algorithms that do not exist.

They added that this assumption is tied to the unresolved P versus NP problem in computer science, and current AI progress is still far from addressing it. At the same time, AI could significantly shorten cryptographic review cycles, with assumptions that once took 10 years to verify now potentially tested by models running for one month.

The experts also said migration work driven by quantum threats has already prepared some systems for AI-related risks, while cryptographic agility can help systems switch methods when weaknesses are found. For developers, they said the most immediate opportunity is using AI to uncover more vulnerabilities before software is deployed.