AI-heavy teams ship vulnerable code at 3.4 times the rate

A Checkmarx report based on 2,350 developers, CISOs, and application-security managers offers a stark picture of AI-assisted software development. Seventy percent of respondents said AI-generated code contains significantly more vulnerabilities, 30% acknowledged knowingly shipping vulnerable code, and 93% reported at least one breach caused by vulnerable applications. Respondents estimated that AI now produces about 49% of production code, slightly below last year's self-reported 54%. The most consequential finding is the relationship between adoption and risk: organizations where AI generates 81% to 100% of code reported shipping vulnerable code at 3.4 times the rate of organizations in the 1% to 20% band. The survey is self-reported, but it captures a real governance gap as coding speed outpaces review and remediation processes.

Key details: 2,350 respondents, 70% report significantly more vulnerabilities, 30% knowingly ship vulnerable code, 93% report breaches, 3.4x vulnerable-code shipping rate at highest AI adoption.

Continue swiping for more AI Brief stories.