Steven Deobald explores the ethical minefield surrounding Large Language Models (LLMs), focusing on plagiarism and the conflict with strict ethical codes like the Buddhist sīla. He argues LLMs inherently violate copyright and encourage dishonesty by consuming pirated content, even if that content includes works by struggling artists. While acknowledging LLMs offer significant benefits, such as translating the Pariyatti app into foreign languages and enabling visually impaired developers to contribute, Deobald highlights a deeper concern: the lies LLMs tell by concealing their vast, unconsented training data.

He warns that using them, even ethically, involves consuming stolen intellectual property. Deobald categorizes developers into cautious 'I've-never-touched-one' types and the experimental 'YOLO' crowd, finding himself in a middle ground where LLMs handle coding minutiae but human oversight remains crucial for architecture and design.