Continuing to review third party sources talking about the new US Copyright Office report about gen AI + copyrightability, before diving in more deeply into the actual document. This one from Ars Technica has a couple interesting points to bounce off of:

“Very few bright-line rules are possible,” the Copyright Office said, with one obvious exception. Because of “insufficient human control over the expressive elements” of resulting works, “if content is entirely generated by AI, it cannot be protected by copyright.”

Curious what “entirely generated by AI” really means here? I know what I think they probably think it means… something like ‘the AI arranged all the pixels’ (?) maybe? But “entirely” to me speaks more to agency, so to suggest an AI system did something “entirely” on its own would mean something more like on its own agency, without human intervention. Which is not at all the case in task based prompting. Maybe if you had a system that was programmed to randomly put together strings of words and use those as prompts, you could make a bit better case for it to be “entirely” done by the system – rather than more directly by a human prompter in real time. But even then, it’s still initiated, designed, and set in motion as a result of human agency. Well, for now.

But this is why I still think the UK copyright system for computer-generated works is just… better (though I know it’s under review currently… not sure current status). I quoted something here in my statement to the USCO from a UK gov office on their legal situation, which has been on the books since the late 1980s, I believe:

“The “author” of a “computer-generated work” (CGW) is defined as “the person by whom the arrangements necessary for the creation of the work are undertaken”. Protection lasts for 50 years from the date the work is made.”

But anyway, my arguments in that area did not prevail.

Back to the Ars Technica piece:

Following Kaminstein’s advice, officials plan to continue reviewing AI disclosures and weighing, on a case-by-case basis, what parts of each work are AI-authored and which parts are human-authored. Any human-authored expressive element can be copyrighted, the office said, but any aspect of the work deemed to have been generated purely by AI cannot.

“Deemed” how, exactly? This will be one thing I am on the lookout for as I begin my deeper dive into the USCO report. Because it may not be altogether very obvious at times – in my books, for example (even I don’t know a lot of the time looking back on some) – where the human ends and the AI begins in creative collaborations with these tools. And unless there is some foolproof method that’s being proposed here… well, we might be back where we started.

“Where a human inputs their own copyrightable work and that work is perceptible in the output, they will be the author of at least that portion of the output,” the guidelines said.

Again “perceptible” sounds a bit wiggly in terms of interpretation. I’m feeling like a lot of these things are not enforceable unless we take works in question into Photoshop and zoom way in at a pixel-by-pixel level to decide which pixels are attributable to a human and which to a machine. That might not be the intent, but it sounds tough to scale for copyright application assessors. For artists, I guess they could start separating out layers of AI versus human actions? Does C2PA or Content Credentials work at this micro-level? Do we have to send our full file history with copyright applications now?