Copyright Office report finds no opt-out right for authors

Here's the headline authors keep repeating to each other: the Copyright Office gave us an opt-out. Here's the problem — it didn't. Read the actual document, Part 3 of the Office's report on copyright and AI, titled Generative AI Training, and you'll find the opposite argument sitting right there in plain language.

The report says requiring copyright owners to opt out is inconsistent with the basic principle that consent is required for uses that fall within your statutory rights. In plain terms: copyright is opt-in. If someone wants to use your work in a way the law reserves to you, they need permission — unless an exception like fair use applies. An opt-out system would flip that burden onto you. The Office isn't recommending that.

There's a conditional worth knowing too. If Congress ever creates a specific exception for AI training, the report says an opt-out could preserve some ability to block unwanted uses or negotiate terms. But even in that hypothetical, the Office flags real design problems: a workable opt-out needs to be simple enough for someone with zero technical expertise, company-by-company opt-outs would be a burden, and a robots.txt-style setting isn't enough — because plenty of owners don't control the sites where their work ends up, including pirate sites.

On the substance of training itself, the report says assembling datasets and training implicate your exclusive rights, and fairness turns on specifics — what works were used, from where, for what purpose. Some training is likely transformative. But training on pirated or illegally accessed works weighs against fair use, and even when outputs don't copy your book outright, they can still dilute the market for books like yours — that counts against fair use too.

None of this is settled law. The report is agency analysis, not a court ruling, and courts haven't followed it uniformly — one judge found training on lawfully acquired books fair use, another granted summary judgment to a developer because the authors hadn't proven market harm. And the report itself is still labeled "Pre-Publication Version" as of this writing, with no date set for a final release.

What to do this week: stop waiting for an opt-out mechanism — none is coming from this report. If you suspect your work was used to train a model, your leverage is proving the use wasn't licensed, wasn't lawfully sourced, or diluted your market — not registering an objection after the fact. Keep records of where your books are available and any evidence of piracy tied to your titles; that's the material a fair-use fight actually runs on.