The official guidance is on SharePoint, and the main new thing is that we should have an attribution in the README. I think we should go a bit further and to per-PR attributions, so as to allow keeping track of what code contains AI generated content. We should probably also record the model used, as this would be relevant for any IP issues that arise.
Describe alternatives you've considered
We could ban GenAI entirely. That would avoid any potential issues with AI generate code, but we would also miss out on any benefits. As the wider corporate strategy at the Met Office seems to be embracing AI, we should probably follow. Also, if AI is banned, it may be used clandestinely, which would make it impossible to track.
I personally have minimal experience with Generative AI development tools currently, so I am not well placed to comment on their effectiveness. We may find that because we attract a particular kind of contribute audience we are less affected by many of the issues affecting the wider open source ecosystem.
What problem does your feature request solve?
With GitHub Copilot now being available at the Met Office, we should have an explicit Generative AI policy.
Describe the solution you'd like
The official guidance is on SharePoint, and the main new thing is that we should have an attribution in the README. I think we should go a bit further and to per-PR attributions, so as to allow keeping track of what code contains AI generated content. We should probably also record the model used, as this would be relevant for any IP issues that arise.
Describe alternatives you've considered
We could ban GenAI entirely. That would avoid any potential issues with AI generate code, but we would also miss out on any benefits. As the wider corporate strategy at the Met Office seems to be embracing AI, we should probably follow. Also, if AI is banned, it may be used clandestinely, which would make it impossible to track.
I personally have minimal experience with Generative AI development tools currently, so I am not well placed to comment on their effectiveness. We may find that because we attract a particular kind of contribute audience we are less affected by many of the issues affecting the wider open source ecosystem.