Even if the generated code is sound, it’s still fundamentally unethical to use LLMs due to the reasons laid out by the anti-LLM proposals (copyright concerns, reviewer burnout, environmental concerns, loss of process knowledge, general security concerns with sending data to untrustworthy service providers, DDOS of open source projects by training data scrapers).
You are welcome to make that call for yourself but Debian is a community project and thus need to be considerate of everyone. It is far to early to understand the long term uses and consequences of AI.
Link to the proposal: https://lists.debian.org/debian-vote/2026/07/msg00284.html
Seems more than reasonable to me, tldr: Use your own judgement, don’t compromise quality.
Even if the generated code is sound, it’s still fundamentally unethical to use LLMs due to the reasons laid out by the anti-LLM proposals (copyright concerns, reviewer burnout, environmental concerns, loss of process knowledge, general security concerns with sending data to untrustworthy service providers, DDOS of open source projects by training data scrapers).
You are welcome to make that call for yourself but Debian is a community project and thus need to be considerate of everyone. It is far to early to understand the long term uses and consequences of AI.