If maintainers of open source want's AI code then they are fully capable of running an agent themselves. If they want to experiment, then again, they are capable of doing that themselves.

What value could a random stranger running an AI agent against some open source code possible provide that the maintainers couldn't do themselves better if they were interested.