Policy

Library and Archives Canada Limits ISBNs for Mostly AI-Generated Books

· 6 min read · Policy

Library and Archives Canada Limits ISBNs for Mostly AI-Generated Books

Library and Archives Canada has tightened the rules behind ISBN Canada, the free service that gives Canadian publishers the International Standard Book Numbers used to identify and distribute an edition, so that books written mostly by artificial intelligence can no longer draw on it freely, Global News reported on October 22. The federal institution updated its guidance on October 9.

Under the revised policy, registering a title amounts to a declaration that it was written by a human author. Writers may still lean on AI for limited support tasks such as brainstorming, grammar checks or formatting, but ISBN Canada now limits how many numbers it issues for works whose text, illustrations or translation are primarily machine-generated. The institution says it is still reviewing its approach and that the guidance may change, and it notes that an ISBN has never conferred copyright protection.

The Canadian Authors Association said the change protects the country's creative sector and lets its cultural identity continue to be shaped by human writers, while calling it only a first step. The association wants to know how self-declarations will be monitored and what penalties apply if a book registered as human-authored turns out to be AI-written. The Association of Canadian Publishers said the policy aligns with its members' values and blamed a flood of low-quality, copycat AI titles sold online for crowding out legitimate books and confusing buyers.

What it means in practice

Library and Archives Canada Limits ISBNs for Mostly AI-Generated Books — contextual photo

For most traditional publishers little changes day to day, since their lists are already human-written. The pressure falls on high-volume self-publishing operations that rely on free Canadian ISBNs to push out machine-written titles, and on retailers and libraries that treat the number as a sign of a real edition. Library and Archives Canada has become a recurring name in Canadian AI stories since Transluce flagged failed AI agent probes of its systems, and the authorship question echoes concerns about AI-generated journalist personas surfacing across Canadian news.

The bigger gap is federal copyright law. Ottawa's AI strategy, released in June, left copyright out, and a consortium of Canadian news publishers moved this month to sue Anthropic over the use of their content to train Claude. Taleeb Noormohamed, parliamentary secretary to AI Minister Evan Solomon, said at the time that work with Culture Minister Marc Miller was ongoing. Global News said Solomon's office had not answered whether the ISBN change is tied to that work.

What to watch: whether Library and Archives Canada publishes a verification process or penalties for false declarations, and whether copyright measures for AI training finally reach the government's agenda.

← Back to AI4Canada