Oct 13, 2026 · a16z Podcast · 5 min read

AI Cryptography Threats and Startup Opportunities

Stanford cryptographer Dan Boneh and A16Z partners analyze the impact of recent AI mathematical breakthroughs on cryptographic security. The discussion clarifies that current systems remain secure, highlights the strategic value of crypto agility, and identifies opportunities for AI-driven software security startups.

Technology Business Entrepreneurship

The takeaways

AI Breakthroughs Do Not Break Crypto. Recent AI mathematical advances have not compromised deployed cryptosystems. Current security assumptions remain valid, and no immediate action is required for end-users or institutions.
Crypto Agility Is Critical Defense. Deploying systems capable of rapid switching between cryptographic primitives is the primary defense against unexpected AI-driven vulnerabilities. This strategy allows organizations to adapt quickly if assumptions fail.
AI Accelerates Security Validation. AI models can stress-test cryptographic assumptions far faster than human experts. This capability allows startups to validate security pre-deployment, significantly reducing the time-to-confidence for new systems.
Hash-Based Signatures Offer Resilience. Hash-based signatures provide greater agility than lattice-based systems because the underlying framework remains sound even if specific hash functions are compromised. This reduces the risk of total system collapse.
AI Code Scanning Reduces Bugs. Integrating AI models into software development cycles enables continuous scanning for vulnerabilities. This approach asymptotically approaches zero bugs, enhancing overall software quality and security.
Lattice Systems Offer Advanced Features. While hash-based systems are more resilient, lattice-based cryptography supports advanced features like threshold signatures. Organizations must balance security resilience with functional capabilities in their architecture.

The AI Cryptography Panic: Fact vs. Fiction

Recent speculation that AI models might break modern cryptography has caused unnecessary alarm. Stanford professor Dan Boneh and A16Z partners clarify that recent AI mathematical breakthroughs have not compromised deployed cryptosystems. Current security assumptions, such as elliptic curve discrete log, remain robust. The absence of cryptographic problems in recent AI research papers suggests that AI models have not yet found efficient attacks on these systems.

Strategic Implications for Security

The core risk is not an immediate break, but the acceleration of the timeline for potential vulnerabilities. AI models can test cryptographic assumptions far faster than human experts. This capability is a double-edged sword: it could reveal weaknesses, but it also allows for rapid validation of new systems. Organizations must adopt crypto agility, ensuring their infrastructure can switch between different cryptographic primitives if one is compromised. This strategy is already being implemented in many internet protocols, which use hybrid encryption combining elliptic curves and lattice-based systems.

Startup Opportunities in AI Security

The convergence of AI and cryptography creates significant opportunities for startups. AI models can be used to scan codebases for vulnerabilities, reducing the number of bugs in software. This capability is becoming a standard practice for companies deploying new cryptographic capabilities. Startups can leverage AI to validate security pre-deployment, providing a competitive advantage in terms of reliability and trust. Additionally, the development of more efficient software scanning methods is a key research area, offering potential for new tools and services.

Conclusion

The current state of AI and cryptography is one of increased confidence, not panic. AI models are powerful tools for both offense and defense. By leveraging AI for security validation and maintaining crypto agility, organizations can protect their systems against future threats. The focus should be on continuous improvement and adaptation, rather than reactive measures. The integration of AI into security practices is a net positive for the digital world, enhancing overall software quality and security.

Key insights

Recent AI mathematical breakthroughs have not compromised deployed cryptosystems. Current security assumptions remain valid, and no immediate action is required for end-users or institutions.

Security Status →

Impact: Reduces unnecessary panic and allows organizations to focus on long-term strategic planning rather than reactive measures.
Crypto agility, the ability to rapidly switch between cryptographic primitives, is the primary defense against unexpected AI-driven vulnerabilities. This strategy allows organizations to adapt quickly if assumptions fail.

Strategic Defense →

Impact: Enhances organizational resilience and reduces the impact of potential cryptographic breaks on business operations.
AI models can stress-test cryptographic assumptions far faster than human experts. This capability allows startups to validate security pre-deployment, significantly reducing the time-to-confidence for new systems.

Operational Efficiency →

Impact: Accelerates product development cycles and reduces the risk of deploying vulnerable systems, providing a competitive advantage.
Hash-based signatures provide greater agility than lattice-based systems because the underlying framework remains sound even if specific hash functions are compromised. This reduces the risk of total system collapse.

Cryptographic Design →

Impact: Offers a more resilient option for organizations prioritizing security over advanced features, particularly in blockchain contexts.
Integrating AI models into software development cycles enables continuous scanning for vulnerabilities. This approach asymptotically approaches zero bugs, enhancing overall software quality and security.

Software Engineering →

Impact: Improves product reliability and reduces the cost of security breaches, making AI-driven code scanning a valuable investment.

Action items

Implement crypto agility in your infrastructure to ensure rapid switching between cryptographic primitives if one is compromised. This strategy is already being implemented in many internet protocols.

Impact: Enhances organizational resilience and reduces the impact of potential cryptographic breaks on business operations.
Leverage AI models to scan codebases for vulnerabilities and validate cryptographic assumptions pre-deployment. This approach can significantly reduce the number of bugs in software.

Impact: Accelerates product development cycles and reduces the risk of deploying vulnerable systems, providing a competitive advantage.
Evaluate the trade-offs between hash-based and lattice-based cryptographic systems based on your specific security and functional requirements. Consider the resilience of hash-based systems and the advanced features of lattice-based systems.

Impact: Ensures that your cryptographic architecture is optimized for both security and functionality, reducing the risk of vulnerabilities and enhancing system capabilities.
Monitor AI research developments for any breakthroughs in cryptographic attacks. Stay informed about the capabilities of AI models in solving mathematical problems related to cryptography.

Impact: Allows for proactive adaptation to new threats and ensures that your security strategies remain current and effective.
Invest in research and development for more efficient software scanning methods. This area offers potential for new tools and services that can enhance security and reduce costs.

Impact: Creates opportunities for innovation and differentiation in the market, while also improving the overall security posture of your organization.

Quotes

“The fact that people try to break them and fail gives us more confidence.”

“We have to move to post-quantum cryptography, which we're already doing.”

“The AIs are good at finding the bugs.”