# Post ideas - 2026-11-16
Based on AI-and-writing news from roughly Nov 9-16, 2026. Figures come from the linked articles and were not independently verified.
## 1. Your memoir has a jurisdiction now
**Angle:** An Ontario court just said OpenAI must answer a copyright suit in Canada, a country with no text-and-data-mining exception, so where an AI model was trained may matter as much to a writer as whether it was.
- On Nov 13, Ontario's Court of Appeal dismissed OpenAI's appeal and kept the news publishers' suit in Ontario. It ruled on venue only, not infringement.
- The publishers (Torstar, Postmedia, Globe and Mail, Canadian Press, CBC) seek up to $20,000 per work over about 10.65 million works.
- Canada's fair dealing is narrower than US fair use, so US-centric "is training fair use?" arguments may not travel.
- What this means for a working writer: copyright protection may look very different depending on the country where your book or blog is read, sold, or scraped.
**Sources:** [TechTimes, Ontario appeal court](https://techtimes.com/articles/330797/20261116/ontario-appeal-court-keeps-openai-copyright-suit-canada-where-us-fair-use-does-not-apply.htm); [Court allows news publishers' suit to go ahead (Canadian Press)](https://www.thecanadianpressnews.ca/science/court-allows-news-publishers-copyright-lawsuit-against-openai-to-go-ahead/article_afdcee71-4320-5549-9702-4c0b595b7764.html)
## 2. If AI training is a market, should writers be in it?
**Angle:** Publishers are arguing in court that a growing licensing market proves training isn't fair use, which turns "should I license my writing?" into a practical question for independent writers and web-project owners.
- On Oct 29, four publishing groups (AAP, CCC, IBPA, STM) filed an amicus brief in the SDNY OpenAI MDL arguing a licensing market weighs against fair use. Replies are due Nov 20.
- They cite an unverified estimate of a roughly $5.2B training-data market in 2025. Ask who actually gets paid: publishers, platforms, or the authors.
- Authors Alliance is seeking to file on the other side, on "market dilution", so even writers' advocates disagree.
- Practical angle for your own sites: opt-out signals, llms.txt (Ghost has supported it since July), and what you'd want in a licence before saying yes.
**Sources:** [Books & Review, publishers' amicus brief](https://www.booksnreview.com/articles/60347/20261104/publishers-openai-ai-training-fair-use-brief.htm); [Authors Guild on the Anthropic settlement](https://authorsguild.org/advocacy/artificial-intelligence/what-authors-need-to-know-about-the-anthropic-settlement/); [Ghost News, Nov 2026 (Mean CEO)](https://blog.mean.ceo/ghost-news-november-2026/)
## 3. Prove you wrote it: drafts are the new detector
**Angle:** When the Caine Prize cleared its 2026 winner using drafts and editing records instead of an AI-detector score, it showed that a writer's messy process is now their best evidence of authorship.
- The prize's internal review reportedly found no sign of generative AI in "The Blue Fighter" and relied on drafts and edit history (as reported by one author newsletter, so worth checking against the prize's own statement).
- Detectors give false positives, and a clean-looking final text proves nothing either way.
- For memoirists: notebooks, voice memos, dated drafts and version history already exist, so a simple habit of keeping them costs almost nothing.
- Possible follow-up: a lightweight "provenance folder" workflow for an Obsidian vault.
**Sources:** [The AI News Authors Actually Need, Nov 14](https://aiweaf.substack.com/p/the-ai-news-authors-actually-need); [Newsweek, grading AI-era writing from drafts](https://www.newsweek.com/universities-ai-writing-assessment-drafts-detectors-students-12718974)