Studies have shown that code generation LLMs produce low-quality output, require high levels of oversight, and have been found to introduce more bugs into code, with security vulnerabilities being a major reason companies have not adopted or have become cold to adopting LLMs.