StarkWare CEO Eli Ben-Sasson recently questioned the disparity between the ability of artificial intelligence to identify software vulnerabilities and its capacity to patch them. In a post on X dated November 18, 2026, he noted that while AI excels at uncovering security flaws, it often fails to provide reliable repairs.
Ben-Sasson highlighted the risks of automated patching, noting that AI-generated fixes can sometimes introduce new, exploitable vulnerabilities. He argued that while bug detection requires broad scanning, effective remediation demands a deep understanding of system architecture to prevent secondary issues.
These comments follow reports that OpenAI successfully used AI to find and patch hundreds of vulnerabilities in the Linux kernel. Ben-Sasson emphasized that discovery and remediation should be evaluated as distinct processes, suggesting that human oversight remains essential for code production.
Earlier this month, on November 5, 2026, Ben-Sasson warned that AI-driven attacks could pose a threat before quantum computing risks materialize. He advocated for the adoption of more resilient cryptographic frameworks, specifically mentioning hash-based systems and ZK-STARKs.
For developers in the blockchain and DeFi sectors, the recommendation is to treat AI as a tool for surfacing potential issues rather than a complete solution for security maintenance. Ben-Sasson suggests that projects should prioritize cryptographic resilience to mitigate future risks.






