NetBSD Foundation presented updated rules making changes to the project source tree. The new rules include a clause prohibiting the inclusion in NetBSD of code generated using AI tools based on large language models, such as ChatGPT, GitHub Copilot and Code Llama, without prior written approval from the Core Team. The reasons cited are copyright uncertainty and non-compliance with NetBSD's licensing policies.
The concern is that AI systems are trained on a large amount of information, including, among other things, code protected by copyright and distributed under different licenses. When generating code by AI systems, these features are not taken into account and potentially the result of AI work can be considered as the creation of a derivative work from the code that was used to train the model and is distributed under certain licenses.
When training a model on code with a license that requires attribution, the code generated by AI tools does not formally comply with this requirement, which can be considered a violation of most open licenses, such as GPL, MIT and Apache. There may also be licensing compatibility issues when inserting code generated using models trained on copyleft-licensed code into projects under permissive licenses such as BSD.
Thanks for reading: