The Local Press Goes to Court, Google Plays Hardball, and Asia Builds Its Own AI

THE LOCAL PRESS FIGHTS BACK

Nearly 400 local newspapers sue OpenAI and Microsoft

The Rundown: The biggest copyright case community papers have ever brought landed in Manhattan federal court this week.

The details:

• A coalition led by Richner Communications filed a 55-page complaint alleging OpenAI and Microsoft copied hundreds of thousands of articles without permission or payment.
• Named plaintiffs include the Arkansas Democrat-Gazette, the New York Amsterdam News, Ogden Newspapers, and Straus Newspapers.
• The complaint alleges both companies 'systematically and secretly crawled' news sites, including pages behind paywalls, then stripped out bylines, publication names, and copyright notices.
• Matthew Platkin, former New Jersey attorney general, is lead counsel; the suit seeks statutory damages, actual damages, return of profits, and legal fees.

Why it matters for us: If you're running a 6,000-circ weekly, this is your fight — the papers suing are your peers, not the New York Times. The outcome could set the price tag for what your archives are actually worth to AI companies.

Read at thenextweb.com →

GOOGLE'S ULTIMATUM

Google tells publishers: share training data or lose Showcase payments

The Rundown: A new AI partnership program comes with strings attached — and a deadline.

The details:

• Google is pitching publishers a pilot program that would promote their content in AI Overviews in exchange for broad rights to use that content for AI training.
• Publishers who decline reportedly face losing payments from Showcase, Google's current content-licensing arrangement, which is being wound down.
• Digital Content Next CEO Jason Kint told the Post: 'There's no fair deal discussions that can happen with Google. It's really a matter of how much money they want to drop.'
• A Pew study found users who see an AI Overview are half as likely to click through to a publisher's site.

Why it matters for us: If your paper relies on Google traffic — and most of us do — the math just changed. You're being asked to trade training rights for visibility you used to get for free. That's a question for your next budget meeting, not a someday hypothetical.

Read at nypost.com →

THE EXPORT BAN FALLOUT

Asian AI labs ship frontier models while Anthropic's Mythos stays locked down

The Rundown: Two weeks into the U.S. export ban, competitors are filling the gap.

The details:

• Chinese firm 360 unveiled Tulongfeng, claiming it matches Anthropic's Mythos, the cybersecurity-focused model banned from export by the Trump administration.
• Tokyo-based Sakana AI launched Fugu, a frontier model it says 'stands shoulder-to-shoulder' with Anthropic's Fable 5 and Mythos Preview.
• Sakana's website now advertises 'delivering frontier capability without the risk of export controls.'
• Sakana co-founder Ren Ito said at the G7 summit that U.S. models 'remain important to Asia' but export controls are reshaping buyer calculations.

Why it matters for us: This feels distant from a county-seat newsroom, but it's not. If AI development fragments by geography, the tools you'll have access to — and the terms you'll get — depend on where you sit. Watch which models your vendors can actually use.

Read at techcrunch.com →

THE UPDATE

Quick hits from the week

The Rundown: Smaller items that shipped, leaked, or surfaced.

The details:

• Reuters reports the White House has released Anthropic's Mythos to some U.S. companies under a restricted pilot — watch for vendor announcements if your AI tools suddenly get more capable.
• The local-press lawsuit names Microsoft Copilot alongside ChatGPT; if your newsroom uses Copilot for Office tasks, you're now a customer of a defendant in a peer lawsuit.

What is?

Tokens

What it is: A token is the chunk an AI actually reads — not a word, not a letter, something in between. Most models break text into pieces of about four characters, so one token is roughly three-quarters of a word. "The" is one token; "newspaper" might be two; a comma is its own token. The AI never sees your sentence as words the way you do — it sees a stream of tokens, and everything it writes back is built one token at a time. You can watch it happen: paste any paragraph into OpenAI's tokenizer and see your own writing split into the pieces the machine reads.

Why publishers care: Two reasons, and both hit close to home. First, it's the meter — every AI tool bills by the token, counting what you feed in plus what it writes back. When a vendor quotes you a price, that's the unit underneath it, which is why a long records dump costs more to process than a short email. Second, it's this week's copyright fight in a single word. When those 400 papers say OpenAI "copied" their articles, what the company actually did was chop every story into tokens and train on them. Your journalism, measured out in tokens, is the raw material — they took the thing in the very same unit they turn around and sell!!

Visit platform.openai.com →