The problem in most of those cases is not specifically AI. Many of the issues you cited are related to Anthropic specifically and many could have been avoided with better testing.
Yes, I am assuming the AI/LLM of choice you've implemented in your software engineering org is Claude because as far as I can tell there aren't really alternatives that come close to its quality in software.
