Cryptographers Assess the Realistic Impact of AI on Cryptographic Security

Cryptographers Assess the Realistic Impact of AI on Cryptographic Security Following recent AI advancements in mathematical problem-solving, cryptographers from Stanford and a16z crypto evaluated whether artificial intelligence poses an immediate threat to modern cryptography, concluding that foundational assumptions remain intact while highlighting AI's potential for defense. Recent breakthroughs in AI math capabilities did not address or weaken specific cryptographic assumptions. Modern cryptography relies on the premise that efficient solutions do not exist for certain problems, meaning an AI cannot discover nonexistent algorithms unless foundational computational complexity questions like P versus NP are resolved. AI can accelerate cryptographic verification by running sustained stress-tests on new assumptions and identifying software bugs prior to deployment. Threats from AI could be broader than quantum risks by targeting post-quantum systems, making cryptographic agility essential. Theoretical Hardness and Mathematical Limits Recent demonstrations of AI solving advanced mathematics did not evaluate cryptography or reveal vulnerabilities in mechanisms such as elliptic curves. Cryptographic security relies on open problems and the fundamental assumption that certain mathematical challenges lack efficient algorithms entirely. Dan Boneh and Justin Thaler point out that if an efficient algorithm does not exist, an AI cannot find it regardless of its capability, and AI has not come close to resolving foundational complexity barriers like the P versus NP problem. Defensive Auditing and Accelerated Confidence Rather than solely threatening systems, AI can significantly compress the traditional multi-year timelines needed to gain confidence in new cryptographic schemes. Researchers can deploy models to continuously attempt attacks against assumptions and examine complex codebases for bugs prior to public release. This lowers the cost and effort required for software shops and startups to eliminate vulnerabilities during development rather than after deployment. Cryptographic Agility and Comparative Threats While quantum computing threats have already motivated transitions away from elliptic curve cryptography, the AI threat is potentially broader because it could also target post-quantum systems. To counter unpredictable advances, experts advocate for cryptographic agility—designing systems that can seamlessly switch between methods, such as existing internet protocols that concurrently apply both elliptic curve and lattice-based encryption.

AI是否正在威胁现代密码学安全？斯坦福与a16z专家拆解技术现实与应对 针对OpenAI等最新数学突破引发的密码学恐慌，来自斯坦福大学和a16z crypto的密码学专家指出，目前的AI进展并未包含针对密码学的结果，也未攻破其底层数学假设。AI不仅不可能找到理论上不存在的高效解法，反而可成为加速密码分析、软件排错和提升密码敏捷性的关键工具。 近期AI在数学与计算机领域的进展并未直接涉及密码学成果，未证明现有椭圆曲线等密码假设安全性降低。 现代密码学依赖于特定数学难题不存在高效算法的假设，且与P versus NP等未解难题相关，AI无法找到不存在的算法。 AI模型能够大幅缩短密码分析验证周期，由以往的等待数年缩短至让模型密集攻击数月以确立安全信心。 虽然AI对后量子密码系统的威胁面可能比量子计算更广，但通过密码敏捷性与AI驱动的漏洞检测能有效抵御风险。 理论基础与算法假设依然稳固 专家指出，当前发布的AI数学突破并未包含破解密码学的内容，并未发现椭圆曲线等特定密码假设出现新漏洞。现代密码学依赖于特定问题没有高效解法的共识，而此类证明与计算机科学核心难题“P versus NP”密切相关。若某难题在理论上本身就不存在高效算法，再强大的AI也无法凭空找出不存在的解法。 AI加速密码学安全验证与缺陷排查 以往一个密码假设提出后往往需要长达十年的人工分析才能投入部署，如今可以利用AI模型运行一个月进行攻击测试，若无法破解则能显著增强行业信心。此外，AI还能作为开发者低成本的防御工具，在复杂的密码软件部署前协助排查逻辑漏洞与实现缺陷。 AI威胁特征与“密码敏捷性”防线 相较于量子计算，AI的潜在威胁面更宽，可能对后量子密码系统展开分析。应对不可预测突破的核心策略是密码敏捷性（Cryptoagility），即让系统能快速切换密码机制。例如当前已有网络连接同时采用椭圆曲线与格密码进行混合加密，即使其中一种被破解也不会导致系统全面失守。

AI is getting better at solving hard math problems — so what does that mean for the cryptography protecting everything from private messages to healthcare to banking and crypto wallets? After OpenAI’s math results sparked some warnings from industry folks that AI could break the cryptography we all rely on, we asked cryptographers from Stanford and a16z crypto what’s fact and what’s assumption, what causes worry and what doesn’t. **Guests:** Dan Boneh — professor at Stanford University who heads the applied cryptography group and co-directs the computer security lab; also special research advisor at a16z crypto Eddy Lazzarin — general partner at a16z crypto Justin Thaler — research partner at a16z crypto and associate professor of computer science at Georgetown University Liz Harkavy — investment partner at a16z crypto Sonal Chokshi — editor in chief at a16z crypto **Highlights:** [00:37](https://www.youtube.com/watch?v=k_CoOoDsZ5Y&t=37s) — What sparked the FUD [01:35](https://www.youtube.com/watch?v=k_CoOoDsZ5Y&t=95s) — What the math can and can’t tell us [05:29](https://www.youtube.com/watch?v=k_CoOoDsZ5Y&t=329s) — Why cryptography depends on open problems [09:25](https://www.youtube.com/watch?v=k_CoOoDsZ5Y&t=565s) — Why the stakes are bigger than blockchains [10:43](https://www.youtube.com/watch?v=k_CoOoDsZ5Y&t=643s) — How AI could actually help strengthen confidence in cryptography [13:50](https://www.youtube.com/watch?v=k_CoOoDsZ5Y&t=830s) — Considerations for users and institutions [20:10](https://www.youtube.com/watch?v=k_CoOoDsZ5Y&t=1210s) — Mapping the threats, AI vs. quantum [23:59](https://www.youtube.com/watch?v=k_CoOoDsZ5Y&t=1439s) — Hash-based versus lattice-based cryptography [33:07](https://www.youtube.com/watch?v=k_CoOoDsZ5Y&t=1987s&pp=0gcJCWMAwfN6Pr3D) — Opportunities for builders and better security tools [38:14](https://www.youtube.com/watch?v=k_CoOoDsZ5Y&t=2294s) — What comes next if there’s a cryptographic break? Plus, six takeaways: Recently published results showed impressive progress across mathematics and computer science, but they didn’t contain cryptography results. “This is incredible progress in mathematics, in computer science, but there are no questions here on cryptography. … So you might wonder, well, does that mean the models are really bad at cryptography? We don’t know.” — Dan Boneh, [03:00](https://www.youtube.com/watch?v=k_CoOoDsZ5Y&t=180s) “We didn’t learn that elliptic curves are vulnerable in a way that they weren’t before. We didn’t learn that any specific cryptographic assumptions have less security than they did before.” — Eddy Lazzarin, [15:19](https://www.youtube.com/watch?v=k_CoOoDsZ5Y&t=919s) Much of modern cryptography rests on the assumption that certain mathematical problems have no efficient solution. This may also be true for increasingly capable AI. “We generally believe there is no efficient algorithm for these problems. It’s not that we think there might be and humans just aren’t smart enough to find it. We think there is not one. And so if there is not one, it doesn’t matter how smart the AI is. It cannot find the nonexistent algorithm.” — Justin Thaler, [16:12](https://www.youtube.com/watch?v=k_CoOoDsZ5Y&t=972s) “We have to assume that certain problems are hard. The reason we can’t prove that those problems are hard is because that’s related to a famous open problem in computer science called the P versus NP problem. “In some sense, we first have to resolve that problem before we can actually argue that certain cryptographic problems really, really are hard. … None of the problems that were posted this week come even close to resolving the P versus NP problem. We still have to keep making assumptions in cryptography, right?” — Dan Boneh, [06:00](https://www.youtube.com/watch?v=k_CoOoDsZ5Y&t=360s) Trying to break cryptography is part of how researchers build confidence in it. A design that survives sustained, serious attacks earns trust. AI could change how quickly that scrutiny happens. “In the past, when someone would make an assumption and build a system based on it, we’d have to wait like 10 years before that system could be deployed — to gain confidence, to give people enough time to analyze the assumption. … And now, all of a sudden, we can accelerate this process tremendously using these models. … Now we can just let a model run for a month, try to break it, and if the model can’t break it, boy, do we have a lot of confidence that the assumption is valid.” — Dan Boneh, [12:10](https://www.youtube.com/watch?v=k_CoOoDsZ5Y&t=730s) While these recent concerns are about AI, some of the same systems are already the subject of migration efforts because of quantum computing. That overlap has limits. “Because of fears of the quantum threat, people have already begun to treat elliptic curve cryptography as something that may be vulnerable soon. So we have already actually engaged in and prepared and deployed many of the mitigations you’d need for those exact concerns for a different threat.” — Eddy Lazzarin, [20:52](https://www.youtube.com/watch?v=k_CoOoDsZ5Y&t=1252s) “The AI threat in some sense is broader, in that it could actually also look at our post-quantum systems and potentially weaken them in some way, maybe break them altogether.” — Dan Boneh, [22:08](https://www.youtube.com/watch?v=k_CoOoDsZ5Y&t=1328s) Cryptographic agility refers to a system’s ability to change its cryptographic methods when evidence or requirements change. It gives engineers a way to respond to discoveries they can’t predict in advance. “We deploy systems that can quickly switch from one type of design to another if the design that’s being used turns out to be insecure. Cryptoagility is in some sense our friend for defending against the unexpected.” — Dan Boneh, [22:47](https://www.youtube.com/watch?v=k_CoOoDsZ5Y&t=1367s) Dan points to internet connections that already combine different cryptographic methods: “They are actually using agile cryptography. They’re encrypted using elliptic curves and lattices at the same time. If you break elliptic curves, it doesn’t matter because it’s also encrypted under a lattice-based system.” — Dan Boneh, [38:40](https://www.youtube.com/watch?v=k_CoOoDsZ5Y&t=2320s) For builders, one of the clearest opportunities is bringing more security testing into development. Discussing complex cryptographic systems, Justin notes: “The surface area for bugs just gets very large, and yes, I’m very optimistic that the same AI tools that are scaring people about cryptography will also help us get all the bugs out. That’s a very hard problem.” — Justin Thaler, [28:56](https://www.youtube.com/watch?v=k_CoOoDsZ5Y&t=1736s) “After it’s deployed, it gets a lot of attention from cryptographers, and that’s kind of the wrong time to attack a cryptosystem. Whereas now, because we control these AI models, any startup or any software shop that’s deploying a new crypto capability or actually even deploying any software at all can now have these models analyze its cryptography and analyze its general software for bugs.” — Dan Boneh, [35:05](https://www.youtube.com/watch?v=k_CoOoDsZ5Y&t=2105s) “Vulnerabilities can be removed. They can be designed away, they can be mitigated, and the primary resource we dedicate to mitigating them is a thing that’s getting radically cheaper every couple months. And that’s AI capabilities.” — Eddy Lazzarin, [37:10](https://www.youtube.com/watch?v=k_CoOoDsZ5Y&t=2230s) *Quotations have been lightly edited for clarity. Ellipses indicate omitted passages.*