Two Sci-Fi Authors Just Explained, in Plain Terms, Why They Won't Touch AI Writing Tools
John Scalzi and Charles Stross aren't making abstract arguments about AI and creativity. They're describing specific, concrete reasons the tools embedded in their own software worry them.
In early August 2026, science fiction authors John Scalzi and Charles Stross went public with detailed, personal objections to AI writing tools — not as a general anti-AI stance, but as specific concerns about what happens to their work once an AI tool touches it.
Scalzi's Provenance Problem
Scalzi's objection centers on the writing tools now embedded directly into the software he already uses. In a July 31, 2026 post on his blog, Whatever, he wrote plainly: "No, Microsoft, I don't fucking want Copilot to 'help me write' in Word, and no, Google, I don't fucking want Gemini to 'help me write' in Google Docs. The very fact these 'AI's are hovering around my writing at all is bad enough from a provenance standpoint."
The concern isn't that the tools might produce bad suggestions. It's that once an AI tool has access to a manuscript in progress, the line between what a human wrote and what a machine touched becomes harder to prove — a problem for any author whose work later becomes part of a copyright dispute.
Stross's Training Data Worry
Charles Stross, one of the named plaintiffs in the Anthropic settlement, raised a related but distinct concern: the risk of unintentionally feeding his own unpublished ideas into systems that could later reproduce them. If he used an AI tool to work through a story outline, he noted, the next version of that model could end up trained on his own unpublished concept — and capable of generating something similar for someone else.
AT A GLANCE • Published: The Register, August 3, 2026 • Scalzi's blog post: Whatever, July 31, 2026 • Scalzi's concern: embedded AI tools (Copilot, Gemini) create provenance problems for his own writing • Stross's concern: sharing story outlines with AI tools risks those ideas being trained into future models • Stross is a named plaintiff in the Anthropic settlement • Comments arrive amid ongoing AI copyright litigation across the publishing industry |
Why Two Established Authors Speaking Out Matters
Neither Scalzi nor Stross is new to this fight. Stross's involvement in the Anthropic settlement means he has direct, material stake in how AI companies are held accountable for training data. Their willingness to describe specific mechanisms — provenance contamination, idea leakage through outline-sharing — gives other authors a concrete vocabulary for objections that often get flattened into general AI skepticism.
Where This Fits
The comments arrive as AI remains, in the words of one industry observer, "the central topic" at publishing and library conferences throughout 2026. As courts work through the Anthropic settlement, the Google lawsuit, and the individual suits from authors like John Carreyrou, working writers are increasingly vocal about what they will and won't allow near their manuscripts — regardless of what courts eventually decide.
Scalzi and Stross aren't arguing AI can't write. They're describing exactly what happens to a manuscript once an AI tool has touched it — and why that's a problem regardless of how good the output is. WriteO News, August 2026 |
WHAT WRITERS CAN TAKE FROM THIS • Provenance is becoming a real, practical concern, not an abstract one — once an AI tool has access to your draft, proving sole authorship gets harder • Sharing unpublished ideas with AI tools, even just to brainstorm structure, carries a real risk of that material ending up in future training data • Established authors going public with specific, mechanical objections — not just general discomfort — gives the rest of the industry clearer language for the same concerns • Software with embedded AI assistants is becoming harder to avoid — knowing how to disable or decline those features is now a practical writing skill, not a preference |
Scalzi's provenance concern points to the same fix regardless of which tools you use. WriteO's Version History keeps a clean, timestamped record of every draft, entirely separate from any AI tool, so your authorship is never in question.
Sources: The Register, August 3, 2026 | Whatever (John Scalzi), July 31, 2026 | Authors Guild