Cybersecurity and software engineering experts issued a stark warning this week: the unchecked reliance on Artificial Intelligence tools for writing code is driving us toward a “software apocalypse.” The report indicates that the speed at which AI generates new lines of code has surpassed human capacity to review, audit, and maintain these systems.

The phenomenon, described as “technical debt on an industrial scale,” suggests that massive codebases are becoming ungovernable. Subtle logic errors and security vulnerabilities, which would slip past simple automated reviews, are accumulating within the critical digital infrastructures of companies and governments.

The Degradation Cycle

According to the analysis, the problem is exacerbated by the fact that new AI models are being trained on code generated by previous AI versions, creating a negative feedback loop that degrades software quality and security. “We are building digital sandcastles without verified foundations,” the study states.

Critical Risks

The saturation of unverified code exponentially increases the attack surface for hackers. Furthermore, the loss of human know-how—with junior developers blindly trusting the machine—could result in a scenario where, in the face of a catastrophic failure, there are not enough qualified experts to fix the problem in time.