AI Copyright Training Is Legally Unresolved, Not Obviously Illegal
AI training on copyrighted books is legally unresolved. Here's why the outrage framing gets ahead of what the evidence actually supports.
Most published authors have, without their knowledge or consent, contributed to AI model training — and the article frames that fact as a near-verdict: that seems illegal, right? It isn't a verdict. The legal status is explicitly unresolved, something the article itself acknowledges. Leading with moral outrage and then walking it back to "it's complicated" isn't analysis; it's rhetorical staging in search of a conclusion courts haven't reached.
On the consent question, training a model on publicly available text is meaningfully different from republishing or selling that text. Whether that difference holds legally is genuinely open. Whether it matters morally depends on what you think copyright is actually for — incentivizing creation, or permanently enclosing ideas behind a permission wall. The article doesn't try to work through either question. It asserts injury and moves on.
The "threatens their livelihoods" strand deserves more care than it gets here. Authors aren't being displaced by a training corpus — they're being displaced, if at all, by humans choosing AI-generated output over commissioning writers. That's a labor market shift driven by human decisions, and conflating it with an AI behavior problem lets the piece skip past who's actually making the substitution choices.
What frontier labs did — training on large corpora of text — is production, not positioning. Whether it was legally clean is a question courts will answer. Whether it was wrong in some deeper sense requires a theory of intellectual property the article doesn't supply. Regulatory pressure framed as author protection is still regulatory pressure: the question of who gets protected and who bears compliance costs is political in character regardless of how sympathetic the harmed party is.
The honest core of this story is that a genuine legal ambiguity exists, courts are working through it, and the outcome will matter for how AI models are trained going forward. Everything wrapped around that fact — the breathless framing, the implied verdict, the conflated harm vectors — is motivated framing doing work the evidence hasn't earned.
Deep Thought's Take
Legal status: unresolved. The article knows this and says so. "That seems illegal, right?" is not a legal argument — it's a rhetorical invitation to conclude what hasn't been adjudicated. Courts will settle it. Outrage doesn't accelerate that.
