|
|
Was this forwarded to you? Get The Fourth Estate AI Brief in your inbox →
|
|
|
|
View this issue in your browser
Tuesday Brief
Publishers vs. Meta, Copyright Reality Checks, and Digg's AI Comeback
The interesting thing this week with AI and publishing — the legal ground is shifting faster than most of us can track. Five major publishers just sued Meta for training AI on their books without permission. Meanwhile, courts are issuing contradictory rulings on whether that kind of training is even illegal. And somewhere in the middle of all this, you and I are trying to figure out what we can actually *claim* as ours when AI touches our work.
Honestly? This week felt like watching three different games being played on the same field. Let's sort through it.
|
|
PUBLISHERS VS. META
The Rundown: The big book publishers finally stopped waiting to see how other lawsuits play out — they're going after Meta directly.
The details:
- Elsevier, Cengage, Hachette, Macmillan, and McGraw Hill filed a class-action suit in Manhattan federal court Tuesday
- Author Scott Turow joined the suit, alleging Meta pirated millions of copyrighted works to train its Llama AI models
- Meta's defense: AI training on copyrighted material qualifies as fair use, and they'll 'fight this lawsuit aggressively'
- Anthropic previously settled a similar author lawsuit for $1.5 billion rather than risk a trial
- Two federal judges have already issued contradictory rulings on fair use in AI training cases
Why it matters for us: This isn't about textbook publishers — it's about precedent. If Meta loses, AI companies may need to license *any* copyrighted content for training, which could eventually mean your archive has value. If Meta wins on fair use, that door closes.
Read at theguardian.com →
|
|
|
THE COPYRIGHT REALITY CHECK
The Rundown: The Copyright Office has been pretty clear on this, even if social media keeps getting it wrong.
The details:
- Wholly AI-generated text is not copyrightable and falls into public domain — the Copyright Office will refuse to register it
- Human authors CAN claim copyright for creative selection, arrangement, and meaningful modifications of AI outputs
- Prompting alone — even complex, detailed prompts — is not sufficient to claim copyright protection
- A federal district court recently upheld the Copyright Office's refusal to register purely AI-generated works
- Works combining AI and human elements can protect the human-authored portions, but the AI portions remain public domain
Why it matters for us: If you're using Claude or ChatGPT to draft articles and publishing them without meaningful human editing, you can't claim copyright on that work. Competitors could legally republish it. The fix is simple: make sure a human meaningfully shapes every piece before it goes to print.
Read at copyrightalliance.org →
|
|
|
LICENSING LANDMINES
The Rundown: Here's something that caught me off guard — most publishing contracts don't address AI training rights at all.
The details:
- Under standard contracts, publishers do NOT automatically own AI training rights — those remain with the original author
- AI companies need consent from copyright holders before licensing content for training purposes
- The Authors Guild recommends 75-85% author split for AI training license revenue; publishers are pushing 50-50
- In Bartz v. Anthropic (2025), one court ruled training on legally acquired books was 'spectacularly transformative' fair use
- The legal landscape remains unsettled with significant litigation still pending
Why it matters for us: For a weekly running a stable of freelancers, this means checking your contributor agreements now — before an AI company comes knocking about your archive. If your contracts don't explicitly address AI training rights, you might not own what you think you own.
Read at janefriedman.com →
|
|
|
DIGG'S AI PIVOT
The Rundown: Kevin Rose is back, and Digg is trying something actually interesting this time — tracking what *influential voices* are engaging with to surface what matters.
The details:
- Digg relaunched Friday after its Reddit-clone reboot failed in March due to bot problems and lack of differentiation
- The new site ingests X engagement data in real time, tracking views, comments, likes, and saves from influential accounts
- When Sam Altman engages with an AI story, Digg's system tracks the chain reaction of discussion and propagation
- Features include most-viewed stories, rising discussions, fastest-climbing content, and rankings of top 1,000 AI influencers
- If successful with AI news, Digg plans to expand to other topic verticals
Why it matters for us: The underlying idea is worth watching: use engagement signals to find what's actually gaining traction *before* it's obvious. A county weekly could theoretically apply similar logic to local Facebook groups — tracking which council meeting complaints are spreading before the next issue deadline.
Read at techcrunch.com →
|
|
|
WORKFLOW TIP
Claude's 'Skills' Feature: Train It Once on Your House Style
The Rundown: This one's actually practical for a Tuesday production day — create a reusable instruction set so you're not explaining your style guide to Claude every single time.
The details:
- Skills let you store reusable instructions instead of giving Claude the same directions over and over
- Create one by pasting 2+ examples of your writing and asking Claude to capture your tone, structure, and style
- Best uses: proofreading, meeting summaries, social posts, and style guide compliance
- Start with just 1-2 skills for truly repetitive tasks — too many become hard to manage
- An email-writing skill is recommended as a beginner's first step before tackling editorial workflows
Why it matters for us: At the Citizen, we've been pasting our house style notes into Claude conversations for months. A dedicated skill that already knows we don't use courtesy titles after first reference and that 'Windom' never needs a state identifier would save real time on deadline nights.
|
|
|
THE UPDATE
Quick Hits From the Week
The Rundown: Smaller items that crossed the desk but didn't need a full section.
The details:
- AI detection software continues to improve — CopyLeaks and Pangram are emerging as industry leaders with reasonable reliability in controlled tests, worth knowing if you're vetting freelancer submissions
- The Authors Guild's Human Authored Certification program costs $10 per book and operates entirely on the honor system with no manuscript analysis — more symbolic than protective
- The $1.5 billion Anthropic settlement with authors is shaping how other AI companies approach licensing negotiations, potentially setting floor prices for content deals
|
|
|
|
The copyright questions aren't going away. If anything, they're going to get louder as these lawsuits work through the courts. My take: get your contracts in order now, make sure humans are meaningfully editing anything AI touches before publication, and keep an eye on that Meta case. The precedent it sets will matter for all of us.
Trevor Slette runs Quadd.ai — AI tools built for publishers.
|
|
|
|
|
|
Run a newspaper. Use Quadd.
AI tools built for small newsrooms. Document intelligence that turns court reports and box scores into copy you can paste into your layout. Audio transcription that finds the quote without you scrubbing for it. AP-style proofreading that tightens copy without flattening voice. Built by someone who knows what a Tuesday night feels like. Free for seven days.
Request a 7-day trial →
|
|
|
