The results of the general vote (GR, general resolution) of the Debian project developers involved in package maintenance and infrastructure support have been published. The issue reviewed was the use of large language models and AI tools in the distribution development. The fifth point, "Responsible Use of Generative AI," won the vote.
The chosen option does not approve or prohibit the use of AI in development, support, documentation creation, and package formation. It acknowledges that responsible use of AI tools can significantly enhance the productivity of participants, freeing them from routine tasks and allowing more time for complex and creative work.
The code created through AI is proposed to be held to the same quality standards and licensing requirements as code written manually. It is not recommended to pass the results of AI work without verification, testing, and understanding the essence of the changes. All responsibility for the changes made through AI lies with the developers who submitted them to the project. Copyright and licensing issues for works created via AI are resolved under the existing Debian policy.
Disclosure of AI usage is recommended but not mandatory. The use of third-party AI services for handling confidential data, such as access keys and credentials, as well as for analyzing vulnerabilities that have not yet been publicly disclosed, is prohibited. Performing mass automated actions using AI requires prior discussion and must be carried out under human supervision.
Source: opennet.ru
