Eli Ben-Sasson questions why AI can find bugs but not fix them
The StarkWare CEO highlights a gap between AI's skill at spotting security flaws and its record on patching them
Artificial intelligence has become very good at finding the cracks in software. Fixing those cracks is turning out to be a different job entirely.
That gap is what Eli Ben-Sasson pointed to in a post on X on November 18, 2026. He asked why AI models can uncover hidden security vulnerabilities but cannot reliably repair them.
The detection-versus-repair problem
Ben-Sasson is the CEO of StarkWare and a co-founder of Zcash. He is best known for his work on zero-knowledge proofs, which let one party prove a statement is true without revealing the underlying data.
His latest comment did not name a specific protocol, company, or incident. It reads as expert commentary on a trend rather than a reaction to a single breach.
The remediation side has not kept pace. Spotting a broken lock is one skill. Installing a new one without leaving the back door open is another.
That distinction matters because a bad patch can be worse than no patch at all. A known bug is at least a known risk. A fix that quietly introduces a fresh flaw creates a risk nobody is tracking.
OpenAI’s Linux kernel work raises the stakes
Ben-Sasson’s remarks arrive alongside discussion of OpenAI’s recent security results. OpenAI reported that its AI pipeline discovered and patched hundreds of vulnerabilities in the Linux kernel.
The concern is in the second half of that sentence. AI-generated patches have been observed to sometimes introduce exploitable vulnerabilities of their own.
AI, tech, and the markets they move—in one daily briefing.
Daily. Free. Join 34,000+ readers across crypto, finance, and policy.
This is where Ben-Sasson’s question gets its edge. Finding bugs rewards breadth: scan everything, flag anything suspicious. Fixing bugs rewards depth: understand the whole system well enough that a change in one place does not cause trouble somewhere else.
A continuation of earlier warnings
This is not the first time Ben-Sasson has flagged AI as a security concern this month. On November 5, 2026, he warned that AI-driven attacks could potentially emerge ahead of quantum threats.
Ben-Sasson’s earlier message was that preparation should not wait. He emphasized moving toward more resilient cryptographic frameworks, including hash-based systems and ZK-STARKs.
ZK-STARKs are the proof technology at the core of StarkWare’s work. Hash-based approaches lean on hash functions rather than the number-theory problems that quantum computers are expected to threaten.
Read together, his two posts sketch a sequence. AI may sharpen attackers before quantum computers arrive, and AI is not yet a dependable defender on the repair side.
What this means for crypto and security teams
For crypto projects, the implications are fairly direct. Blockchains, DeFi protocols, and identity systems run on cryptography and on code that often controls real money.
That suggests a practical takeaway for development teams. AI can be a useful tool for surfacing problems, but its proposed fixes likely still need careful human review before they go anywhere near production.
Ben-Sasson’s broader push also points to where security spending could flow. Sectors that depend heavily on cryptography, such as DeFi and identity, may see more interest in hash-based systems and technologies like ZK-STARKs if the emphasis on resilience keeps growing.
It is worth noting that Ben-Sasson leads a company built on one of the technologies he is recommending, so readers can weigh his advocacy with that context in mind.
For now, the signal from Ben-Sasson is that discovery and remediation should be judged separately. The headlines about AI finding hundreds of bugs are only half the scorecard.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.

