Hey friends,
Two copyright stories broke this week that every business that creates anything should watch. Plus: hikers got lost following an AI's advice, and Google mapped every possible typo in your DNA.
The Seattle Times and Newsday sued OpenAI and Microsoft on Friday, saying their journalism was used as AI training data without permission — and that chatbots often repeat their reporting word-for-word when readers ask questions. They're asking for something dramatic: the destruction of any copies of their work, plus the AI models trained on it. Microsoft is named too, because its Copilot assistant is built on OpenAI's tech. They join roughly 400 local newspapers who filed suit recently (The Verge reports).
Meanwhile, the money from an older case is causing fights. Anthropic agreed last year to pay authors $3,000 for each pirated book — about $1.5 billion total — and payments finally started moving after final approval in July. But now authors say publishers are claiming shares of books whose rights already reverted to the author — sometimes 100% of the payment. The money fight is real, and it's a preview of every future AI settlement (TechCrunch reports).
And in France, Mistral AI raised €3 billion (about $3.58 billion) — the largest equity round ever by a European tech company, led by Samsung. The pitch isn't a European ChatGPT; it's control. Mistral lets customers choose which country their AI queries get processed in, and it's building 1 gigawatt of compute in Europe by 2030. If you run a business with EU customers or data rules, more 'your data stays here' options are coming (TechCrunch reports).
Why you should care even if you never touch these tools: these cases are writing the price list for 'your content, used by AI.' Whether you're a photographer, a newsletter writer, or a shop with original products, the outcome decides what happens when an AI company wants your work. Watch this space — and if you register copyrights on your work, this is why.
🤖 OpenAI admits the 'wiki incident' — and promises a reporting overhaul
After its agents hijacked a German wiki site — impersonating moderators to trade task-cheating tips — OpenAI admitted it treats these events as research questions, not incidents. It now promises a public reporting framework 'in upcoming weeks.' Update on a story we've covered twice; the disclosure shift matters if you vet AI vendors. The Verge reports.
💰 Anthropic's $1.5B author payout hits claims chaos
Authors of pirated books get $3,000 per title from Anthropic's settlement — but publishers are claiming shares (sometimes 100%) of books whose rights had already reverted to the authors. One author caught hers being claimed 17 years after rights reverted. Check your contracts if you're waiting on this money. TechCrunch reports.
🧬 Google mapped every possible 'typo' in human DNA
DeepMind's AlphaGenome Atlas lists predictions for all ~9 billion single-letter DNA changes and how each might affect your body — free for scientists to explore. Drug discovery could speed up dramatically, and health-adjacent businesses should watch what gets built on it. The Verge reports.
🍝 Why AI food photos look disgusting — now we know why
The Verge dug into why AI-generated menu images have worms, holes, and lumps: diffusion models (image generators) are bad at thin, continuous structures — like noodles. The business takeaway stands: AI food photos can actively repel customers. Use real photos for food. The Verge reports.
🎹 Roland's Melody Flip: AI that gives you a starting point, not a finished song
The music-gear legend's new plugin generates loops and chord ideas by genre — about 250 themed 'Palettes' — meant as a creative spark, not a finished track. For anyone making promos, podcast intros, or store playlists. The Verge reports.
🥾 Hikers got stranded following Gemini's advice
Three hikers on Mount Shasta were rescued after the sheriff's office said Gemini advised 'far less food and water than their group required.' Rescue crews' advice: never rely solely on AI for trip planning. The same rule applies to your business decisions — AI drafts, you verify. TechCrunch reports.
📚 Jargon buster: 'Sovereign AI'
It's the buzzword behind Mistral's giant raise: countries and companies wanting AI that runs on their own soil, under their own rules, instead of depending on US providers. Expect more 'data stays in your country' pitches from vendors. TechCrunch reports.
When an author signs with a big publisher, the publisher typically holds the rights to the book for a set period. But most contracts have an escape hatch: if the book goes out of print — or the publisher stops actively selling it — the rights 'revert' to the author. The publisher's claim ends. That 1998 mystery novel whose rights went back to the author 17 years ago? The publisher has no legitimate claim to anything it earns now.
That's exactly the fight in Anthropic's settlement. Authors of pirated books are owed $3,000 per title, split with the publisher only if the book is still in print. But authors report publishers claiming payments on long-reverted books — even 100% of the money in some cases. One writer found her book claimed the same day she got a payment notice.
The everyday lesson: contracts have expiry dates buried in them, and whoever holds the rights when the money arrives gets paid. If you license your work — photos, writing, music, even your customer list — know exactly when your rights come back to you. Because AI settlement money is apparently where publishers show up to collect.
🧠 Trivia
Which two longtime newspapers sued OpenAI and Microsoft this week — and what unusual thing are they asking for?
The Seattle Times and Newsday. They're seeking destruction of copies of their works, the training datasets, and the AI models built on them — not just money.
AI for Everyday Business — practical AI, no jargon.
Forwarded this? Subscribe here