What are the key takeaways from this a16z Podcast episode?

Key insights include: **Cryptographic assumptions vs. proven security:** All modern cryptography rests on unproven hardness assumptions—problems like elliptic curve discrete log and shortest vector in lattices—because resolving their true difficulty requires solving the P vs. NP problem first. No progress from OpenAI's 700-problem mathematical release touched any cryptographic problem, theoretical or applied, making the absence conspicuous but not conclusive.; **AI as accelerated stress-testing tool:** Frontier labs could now compress a decade of cryptographic confidence-building into roughly one month by running their most capable models against standardized assumptions. Even a partial result—say, shaving 10 bits of security from a deployed scheme—would constitute actionable intelligence, giving practitioners more precise calibration than years of conventional academic scrutiny currently provide.; **Crypto agility is the primary defense:** Systems designed to swap underlying cryptographic primitives without full rebuilds—a practice called crypto agility, already over a decade old—are the correct structural response to any assumption break. Many TLS connections today already encrypt simultaneously under elliptic curves and lattices (X25519 + ML-KEM-768), meaning a single-assumption break does not compromise the session.

What did Dan Bonnet and Eddie Lazarin discuss on a16z Podcast?

Stanford cryptographer Dan Boneh joins a16z's Eddie Lazarin and Justin Thaler to assess whether OpenAI's recent mathematical breakthroughs—spanning roughly 700 problems across computer science—pose a genuine threat to elliptic curve cryptography securing banking systems, private messages, and blockchain wallets, and what institutions should actually do right now. Key topics include: **Cryptographic assumptions vs. proven security:** All modern cryptography rests on unproven hardness assumptions—problems like elliptic curve discrete log and shortest vector in lattices—because resolving their true difficulty requires solving the P vs. NP problem first. No progress from OpenAI's 700-problem mathematical release touched any cryptographic problem, theoretical or applied, making the absence conspicuous but not conclusive.; **AI as accelerated stress-testing tool:** Frontier labs could now compress a decade of cryptographic confidence-building into roughly one month by running their most capable models against standardized assumptions. Even a partial result—say, shaving 10 bits of security from a deployed scheme—would constitute actionable intelligence, giving practitioners more precise calibration than years of conventional academic scrutiny currently provide..

How long is this episode of a16z Podcast?

This episode is 42 minutes long. SignalCast provides an AI-generated summary so you can get the key insights in about 3 minutes.

a16z Podcast

Can AI Break Cryptography? What the Experts Actually Think

October 13, 2026

42 min episode · 2 min read

Dan Bonnet,Eddie Lazarin,Justin Thaler

Episode

42 min

Read time

2 min

Topics

Fundraising & VC, Design & UX, Artificial Intelligence

AI-Generated Summary

Published Oct 13, 2026

Key Takeaways

✓Cryptographic assumptions vs. proven security: All modern cryptography rests on unproven hardness assumptions—problems like elliptic curve discrete log and shortest vector in lattices—because resolving their true difficulty requires solving the P vs. NP problem first. No progress from OpenAI's 700-problem mathematical release touched any cryptographic problem, theoretical or applied, making the absence conspicuous but not conclusive.
✓AI as accelerated stress-testing tool: Frontier labs could now compress a decade of cryptographic confidence-building into roughly one month by running their most capable models against standardized assumptions. Even a partial result—say, shaving 10 bits of security from a deployed scheme—would constitute actionable intelligence, giving practitioners more precise calibration than years of conventional academic scrutiny currently provide.
✓Crypto agility is the primary defense: Systems designed to swap underlying cryptographic primitives without full rebuilds—a practice called crypto agility, already over a decade old—are the correct structural response to any assumption break. Many TLS connections today already encrypt simultaneously under elliptic curves and lattices (X25519 + ML-KEM-768), meaning a single-assumption break does not compromise the session.
✓Hash-based vs. lattice-based signatures trade-off: Hash-based signatures offer a modular security model—swap one hash function, the framework remains sound—while lattice-based schemes enable threshold signatures, non-hardened key derivation, and richer algebraic structures unavailable with hashes. Blockchain developers choosing exclusively hash-based signatures reduce assumption count but lose significant cryptographic capability and introduce larger, stateful signature overhead.
✓AI improves defensive security more than offensive: Because the number of software bugs is finite and AI inference costs have dropped roughly 95–99% within a year, companies already scan codebases with each new model release and each software update. This asymptotically reduces exploitable vulnerabilities, meaning AI's net effect on global security is likely positive rather than erosive, provided responsible disclosure protocols govern any discovered breaks.

What It Covers

Stanford cryptographer Dan Boneh joins a16z's Eddie Lazarin and Justin Thaler to assess whether OpenAI's recent mathematical breakthroughs—spanning roughly 700 problems across computer science—pose a genuine threat to elliptic curve cryptography securing banking systems, private messages, and blockchain wallets, and what institutions should actually do right now.

Key Questions Answered

•Cryptographic assumptions vs. proven security: All modern cryptography rests on unproven hardness assumptions—problems like elliptic curve discrete log and shortest vector in lattices—because resolving their true difficulty requires solving the P vs. NP problem first. No progress from OpenAI's 700-problem mathematical release touched any cryptographic problem, theoretical or applied, making the absence conspicuous but not conclusive.
•AI as accelerated stress-testing tool: Frontier labs could now compress a decade of cryptographic confidence-building into roughly one month by running their most capable models against standardized assumptions. Even a partial result—say, shaving 10 bits of security from a deployed scheme—would constitute actionable intelligence, giving practitioners more precise calibration than years of conventional academic scrutiny currently provide.
•Crypto agility is the primary defense: Systems designed to swap underlying cryptographic primitives without full rebuilds—a practice called crypto agility, already over a decade old—are the correct structural response to any assumption break. Many TLS connections today already encrypt simultaneously under elliptic curves and lattices (X25519 + ML-KEM-768), meaning a single-assumption break does not compromise the session.
•Hash-based vs. lattice-based signatures trade-off: Hash-based signatures offer a modular security model—swap one hash function, the framework remains sound—while lattice-based schemes enable threshold signatures, non-hardened key derivation, and richer algebraic structures unavailable with hashes. Blockchain developers choosing exclusively hash-based signatures reduce assumption count but lose significant cryptographic capability and introduce larger, stateful signature overhead.
•AI improves defensive security more than offensive: Because the number of software bugs is finite and AI inference costs have dropped roughly 95–99% within a year, companies already scan codebases with each new model release and each software update. This asymptotically reduces exploitable vulnerabilities, meaning AI's net effect on global security is likely positive rather than erosive, provided responsible disclosure protocols govern any discovered breaks.

Notable Moment

Boneh points out that if a frontier lab spent tens of millions of dollars running its most capable model against a cryptographic scheme and failed to break it, that single result would provide stronger security confidence than anything the conventional academic research process could generate over an entire decade.

Know someone who'd find this useful?

Episode Transcript

All of a sudden, a new threat emerged in the world of cryptography, which is whether our AI friends can actually attack cryptosystems that we thought were secure. The fear here is that the AI labs could break them if they are breakable. The fact that people try to break them and fail gives us more confidence. NIST is already standardizing many other types of public key systems that we can switch to. So if it turns out one of our assumptions is broken, we have backups. We just need to know. Are there startup opportunities here and where a new company might be able to come in and help? AI is getting better at solving difficult math problems. Could it eventually break the cryptography protecting our banking systems, private messages, and crypto wallets? Recent results from OpenAI have sparked new concerns about the security of modern cryptography. In this episode, a16z Crypto's Liz Harkabee is joined by Stanford professor Dan Bonnet, general partner Eddie Lazarin, and research partner Justin Baylor to separate fact from speculation. They discuss what the recent breakthroughs actually tell us, how AI threats differ from quantum computing, and why there's no reason to panic today. They also explained how AI could ultimately help make cryptography and software more secure. Hey. Welcome to our our show today. I'm Liz. I'm joined here with Dan Bonnet, a professor at Stanford, and Eddie Lazarin and Justin Thaler, two members of our team here at e sixteen z. We're here to talk about a a pretty big discussion that's going on right now that's been kicked off yesterday by a tweet that a security researcher, Justin Drake, put out warning the crypto industry to start preparing for what he's calling bunker mode, specifically around this worry that the advances in AI are gonna cause a breakthrough or breaking of cryptography faster even than a quantum computer might, which is normally the basis of the discussion that we have when we think about modern day cryptography and and what the state of the art looks like. And so he called on, you know, big institutions to start thinking about moving funds and start thinking about this. And a lot of that has been kicked off by some papers that OpenAI put out and different kind of breakthroughs they had in mathematical problems that were previously unsolved and how that relates back to cryptography. And so, Dan, I don't know if you wanna kinda summarize what those breakthroughs were or what those papers put out and how that kind of kicked this off. So let's see. I guess the first thing I should say is, you know, everybody calm down. Don't panic. Calm down. The world today is the same as it was a year a year ago, two years ago. That's the first message. Now I guess before we get started, I just want to make sure people are on the same page, and that there are kind of two …

Get the full transcript (7,464 words) + summary by email — free

One-time email with the complete transcript and AI summary of this episode. No account needed.

One email, no spam. We’ll also show you what SignalCast does.

Browse all a16z Podcast transcripts →