NEWS
What happened, what it means, and what to do about it.
Latest

Meta's AI can rewrite your ad. The new terms say you own the consequences.
Meta's revised commercial terms, reported to take effect Oct. 30, 2026, make advertisers responsible for anything Meta's AI creates or edits in a campaign — including features turned on by default. Here's what to verify before the deadline.
ALL STORIES

A startup wants you to sell "AI skills" instead of a newsletter. Read the storefront terms first.
Skillfully launched Oct. 29 saying authors and experts can build no-code agent skills and sell them as subscriptions that run inside ChatGPT and Claude. Pricing, payout split and official distribution are not confirmed. Here's what to verify before you build.

OpenAI's GPT-6.1 Terra can operate a computer. The reported price is the part to check first.
OpenAI launched GPT-6.1 Terra on Oct. 26 with computer use and visual reasoning, available in ChatGPT Work, Codex and the API. Reports put the standard rate at $1 per million input tokens and $5 per million output. Token cost is not the bill.

Meta deleted the word "train" from its new terms. The crawl permission stayed.
Meta pushed its updated business terms from Oct. 30 to Nov. 20 and cut the word "train" from the draft. The language letting it crawl sites tied to a Meta Business account is still there, and there is still no compensation.

Google Cloud now sells both OpenAI and Anthropic models. Price the same job three ways before you commit
Google Cloud added OpenAI models in private preview on Oct. 21 and Anthropic's Claude Platform on Oct. 22. One invoice, two rival vendors. What nobody has published yet is the price.

Spotify wants podcasters to sell memberships. Here's what the 10% cut means
Spotify added paid memberships for podcast creators, with a reported 10% platform cut. The pitch is recurring revenue instead of ad sales. The terms you actually need — fees, eligibility, payout mechanics — are not public yet.

Spotify says AI-flagged tracks are 10x more likely to miss its royalty threshold
Spotify is pushing AI disclosure into song credits and keeping AI personas out of recommendations by default. The money part is quieter: tracks it reads as probably AI-generated rarely clear 1,000 streams.

Meta delayed its new publisher terms after a backlash. The delay is the only confirmed fact.
Meta postponed terms that trade groups say would hand it journalism for AI products without payment. No new date, no opt-out, no text. Here's what publishers should do with the pause.

The monetization line on AI video isn't the label. It's repetition.
YouTube and TikTok are tightening rules on mass-produced synthetic video without banning AI-assisted work. The test is originality per upload, not whether AI touched the file.

Meta's new AI terms have publishers at regulators' door. Check your account terms this week.
Publishers asked Brussels and London on October 21 to block Meta's new AI terms, which reportedly reach any business running a Facebook page, an Instagram account, or ads. The exact permissions have not been published.

Forbes joined Google's paid AI news pilot. The data is the part worth copying.
Google's pilot pays publishers for expanded rights to use their journalism in AI overviews and audio briefings. Forbes says the citation and performance data is what it's actually after. Terms were not disclosed.

Does YouTube make you disclose AI music? The label's rules, and the two gaps in them
YouTube started applying a "Made with AI" label to tracks on October 14 and says it will expand beyond YouTube Music. Disclosure sits with the uploader — and clearing a song is still a separate job from labeling it.

Can you still monetize AI video? Labeling it is not the same as clearing it
YouTube, TikTok and Instagram are tightening how AI-assisted content gets labeled, recommended and paid. The rules target volume, not AI itself — and a disclosure checkbox does not make a video eligible for revenue.

YouTube is labeling AI-generated music. What that changes for your soundtrack
YouTube Music began showing labels in October 2026 for tracks made primarily with generative AI, including AI-assisted composition and vocals. The label describes how a track was made — it does not clear you to use it.

Meta's new open model is cheap to run. The license puts Meta's name on your product.
Muse Spark Open shipped October 15 with downloadable weights and hosted access from $1.25 per million input tokens. Anything you build with it has to display "Built with Muse Spark Open" — and model-as-a-service companies over $20M need a separate deal with Meta.

Meta Is Blocking ByteDance Ads in the US. What Publishers and Media Buyers Should Check This Week
Meta started restricting advertising from TikTok's parent company on October 8 across seven markets, including the US. What's confirmed is the ByteDance block. What isn't confirmed is whether your TikTok-destination campaign still runs.

Meta's Muse Wonder can drive a browser. The best version sits behind a $200 plan.
Meta launched an agentic model that reportedly operates browsers and desktop apps. The capability story matters less than the pricing story: advanced AI labor is now a tiered software line item, from $20 to $200 a month.

Do you have to label AI video? The YouTube, TikTok and Meta rules, side by side
Three platforms, three wordings, one shared trigger: if a synthetic person, voice, place or event looks real, it needs a label. Here's what each rule covers, what it doesn't, and where creators get caught.

Claude Haiku 5.5 cut small-model API prices by 90%. The catch is in the long prompts.
Anthropic's cheap model now lists at $0.10 per million input tokens. That makes bulk transcript and metadata work nearly free — until your prompt crosses 100,000 tokens and the price multiplies by five.

USA Today sued OpenAI for more than $250 million. That number is a demand, not a price.
One of America's largest news publishers filed in Manhattan federal court over training on hundreds of thousands of articles. Here's what it changes for anyone negotiating an archive deal — and what it doesn't.

An AI licensing deal finally published a price: $1 per 1,000 tokens
Six academic publishers put 20 million papers behind a pay-per-use meter with Redpine. It's the first public rate card in AI licensing — and the first time you can do the math on your own archive.

Trump announced a federal AI task force. None of your state compliance work changes yet.
ABC7 reported the White House is creating a "Super Intelligence Force" AI task force. The text isn't public, preemption is unconfirmed, and the state deepfake rules you built around are still the only ones with teeth — such as they are.

Publishers built the meter for AI licensing. No AI company has plugged it in.
SPUR, backed by the BBC, Financial Times, Guardian, Telegraph, Sky News and the Associated Press, published an open standard to track five stages of AI content use. Not one AI company has joined its advisory board.

A judge just ruled Google never promised you traffic. That closes the antitrust route.
Judge Amit Mehta dismissed Penske Media and Chegg's antitrust suits over AI Overviews. The reason wasn't that the harm is fake — it's that publishers let Google crawl for free, voluntarily, for decades.

Is it legal to train AI on your work? An appeals court answered for the first time — with an asterisk
The Third Circuit affirmed that ROSS Intelligence can't claim fair use for training on Westlaw headnotes. It's the first US appellate ruling on AI training and fair use. It also explicitly isn't about generative AI.

The going rate for a book in an AI training deal is $5,000 — and your publisher may split it with you
Book publishers spent three years suing AI companies. Now Google is signing licenses with Bloomsbury, De Gruyter Brill and Johns Hopkins University Press. The only disclosed per-title price in the market is still the HarperCollins-Microsoft number: $5,000, three years, 50/50 with the author.

Internal Bing data says Copilot cut publisher click-throughs by up to 94%
A number pulled from the OpenAI-Microsoft copyright litigation puts a figure on what AI answers do to outbound clicks. "Up to 94%" is the ceiling, not the average — and that distinction decides whether you should replan revenue this quarter.

Meta's Oversight Board wants AI deepfakes removed and monetization tightened. Only part of that binds Meta.
The board is pushing Meta toward faster deepfake removals and stricter rules on who can earn from synthetic content. Case rulings bind the company. Policy recommendations do not. Creators should plan for the strict version anyway.

Google's AI payment program covers 200-plus titles worldwide. Do the math before you replan revenue.
Digiday puts Google's broader news AI scheme at more than 200 titles globally. That is the whole addressable pool right now — and the reason the pilot should not appear in your 2027 budget yet.

OpenAI and Microsoft executives called their own products "largely substitutive" for news. It's in the court record now.
Filings made public September 17 quote OpenAI and Microsoft leaders describing their tools as an "existential threat" to publishers. That language targets the exact fair-use factor the companies need to win — and it resets what your archive is worth.

Media execs say they want AI deals. They're rejecting the terms on offer.
Axios reported executives sounding the alarm on AI the same week Google's publisher payment pilot surfaced. The tension isn't whether to license — it's who sets the price, and almost nobody outside the biggest publishers is in the room.

Google's AI licensing pilot pays per value, not a lump sum. That one word decides who gets a check.
Google is testing payments to US publishers when their work materially feeds answers in Gemini, AI Overviews and AI Mode. The model is described as pay-per-value rather than a flat fee — which means most sites in the pilot get a variable number, not a guaranteed one. The formula is still unpublished.

Cloudflare switched on crawler blocking by default. Ad-supported pages are the test case.
As of September 15, Cloudflare blocks AI crawlers that feed training and agent retrieval on ad-supported pages by default, and its pay-per-crawl experiment is now framed as Pay Per Use — payment when your content shows up in an answer, not when a bot fetches the page. The mechanics are thinner than the announcement.

The top AI labs spent the week saying "slow down." Four of them shipped models the week before.
Reuters reported Sept. 12 that Anthropic's CEO called on AI companies to slow model development, and Musk and Altman backed the framing. OpenAI is also pressing Congress for binding frontier rules. Here's what actually changes for people who publish — and what doesn't.

California's AI and deepfake package is law. The federal version is still a draft.
California enacted a package covering companion chatbots, AI-generated child exploitation material and age signals, with penalties reported as high as $250,000 per action for some deepfake conduct. In Washington, Reuters says Senate negotiators are still arguing over a pre-release safety duty. Here's the part that touches your channel.

Google's publisher payments run through Search Console and cover three AI surfaces
Digiday reports Google is rolling out a pay-per-value AI licensing program that pays when publisher content significantly contributes to answers in Gemini, AI Overviews and AI Mode. The delivery channel is Search Console. The payout formula is still not public.

Publishers spent $113M in one month buying back traffic Google used to send for free
A new tally says news publishers poured $113 million into paid search in a single month to recover clicks that organic results used to deliver. Meanwhile Google is piloting payments for content that feeds AI answers. One line is running way ahead of the other.

Two more regional papers sued OpenAI and Microsoft. Same week Google started paying publishers.
The Seattle Times and Newsday filed copyright claims against OpenAI and Microsoft, extending a fight that now runs on two tracks at once: litigation over archives and a Google pilot that pays for live use in answers. The two tracks price your work very differently.

Google is quietly paying publishers when their work feeds AI answers. Nobody has seen the formula.
Digiday reported on Sept. 14 that Google is scaling a Search Console-based program that compensates publishers whose content contributes significantly to Gemini, AI Overviews and AI Mode answers. No payout rates, no publisher list, no launch scope confirmed.

The DOJ told a judge to rule for OpenAI in the New York Times case. Your archive's price is on the line.
The Washington Post reported on Sept. 2 that the Justice Department urged the judge in NYT v. Microsoft and OpenAI to rule for the AI companies. No decision yet — but the fair-use question sets the floor under every licensing deal a publisher signs.

Publishers are restaffing for a world without Google referrals. SEO volume is the first line item to go.
Digiday reports that newsrooms are rebuilding their org charts for the post-Google era. The practical read for independent creators: search-optimized commodity output is the budget getting cut, and direct-audience work is what's getting hired.

Suno shipped label-licensed music models. That doesn't mean your video is cleared.
Suno launched new models built with major-label licensing, per Variety and Gizmodo. The question for creators isn't whether the songs sound good — it's whether the license survives Content ID.

AT&T says open models cut its AI bill by up to 80%. Here's what that means for your stack.
The New York Times reports enterprises are shifting workloads from closed frontier models to open-weight ones, with AT&T claiming savings of up to 80%. For publishers, the lesson is not "switch models" — it's "stop paying frontier prices for chores."
Get the stories by email
What changes in AI for people who publish — sourced, and to the point.
No spam. Unsubscribe anytime.