One in five security breaches now thought to be caused by AI-written code (www.techradar.com)

Aikido Security’s State of AI in Security & Development report finds AI-written code is already a major source of risk: 69% of organizations have discovered vulnerabilities in AI-generated code while AI now produces about 24% of production code globally. The study highlights a governance gap—security teams (53%), developers (45%) and even mergers (42%) are blamed when AI code fails—creating uncertainty over who owns remediation and liability. Aikido’s CISO Mike Wilkes warns this diffusion of responsibility makes breaches harder to trace and fix. The report also ties higher incident rates and slower fixes to tool sprawl: organizations using six to eight AI tools saw incidents at a 90% rate versus 64% for those using one to two, and median remediation stretched from 3.3 days (1–2 tools) to 7.8 days (5+ tools). Regionally, the U.S. reported more serious incidents (43%) than Europe (20%), partly because U.S. developers are likelier to bypass security controls (72% vs 61%) and European firms face stricter compliance. Despite risks, optimism remains: 96% expect AI to produce secure, reliable code within five years and 90% think AI can handle penetration testing in ~5.5 years—though only 21% believe this will occur without ongoing human oversight.