When AI-Generated C# Slips Through Review and Breaks Production

As AI-generated code becomes more prevalent in pull requests, a critical vulnerability is emerging: code that passes tests, clears code review, and appears architecturally sound still introduces production failures. These regressions—missing AsNoTracking() on read queries, incorrect dependency lifetimes, dropped CancellationTokens, and swallowed exceptions—reflect the AI's lack of team context and hard-won conventions rather than syntax errors, making them invisible to both reviewers and automated analyzers.