I was reading articles about Public Domain Day when this idea began to take shape.

The commentary about Public Domain Day leans into excoriating the regressiveness of our copyright regime which may have been reasonable at one time – the Founding Fathers specified a 14-year copyright (if you applied), extendable for another 14 years (if you re-applied) – but which corporate rent-seekers have lobbied Congress to expand into our current automatically-granted-lifetime-plus-70-years travesty. Member of the Open Culture community agree that this assault on the public domain should be rolled back.

On the other hand, when the subject of AI training data comes up, the members of the Open Culture community seem to agree that the idea of large corporations training their Large Language Models on copyrighted data is reprehensible, and we should enjoin these companies from trespassing on our data.

So, it seems to me that a compromise that would make everyone happy is that we should roll back copyright protections to only works that have applied for it – and the copyright period should be 14 years, as our Founding Fathers intended. Then the Large Language Models can be trained on all text published before 2010 – as well as all writings since then which haven't applied for copyright protection (which would be most of what's available on the Internet, I would imagine). I haven't applied for copyright on this post, for example.

Of course, since the grant of copyright was automatic until now, we should definitely provide a transitional grace period within which all works published since 2010 could apply for copyright, should they feel it was needed.

This proposal could easily get the backing of all the people who stand to profit from AI – which seems like all the major technology players. As well as the Open Culture community. They might be able to bring enough pressure to bear to roll back our oppressive copyright regime.

Right?