Robert C. Martin now skips reading LLM-generated code and relies on tests and constraints. Code reviews by humans also miss defects. Software engineering may work without human code review. Teams of AI agents can write and test code. Errors will still occur in LLM-driven releases. Serious defects can cause major financial or safety incidents. Liability for such errors may still fall on humans. Courts may hold individuals responsible regardless of AI involvement. Organizations must consider legal risks before adopting full AI-driven development. Drivers may over-rely on AI systems despite legal responsibility. Legislation often lags behind technological change.