The_Ledger

For years, telling viewers that a video was touched by AI was left to the creator. YouTube has now rewritten that guidance: it will add an AI label on its own when it detects its own AI tools, a C2PA provenance tag, or a match from its internal systems, and several of those labels cannot be removed. The significance is not that YouTube added another label. Its that provenance metadata is beginning to trigger platform behavior automatically.

In human terms: Say you blend in a realistic AI background and post the result. Even if you forget to check the disclosure box, YouTube may detect the AI and label the video for you. If the file carries an industry “made with AI” tag (a standard called C2PA) or you used YouTube’s own AI tools, that label is locked. You cannot remove it, and repeatedly skipping disclosure can cost you content or your place in the YouTube Partner Program.

Why this matters: For creators, “disclosed AI use” is moving from something you control to something the platform decides and enforces. That changes how AI work appears to your audience and to advertisers, and it ties your channel’s standing to whether YouTube’s detection agrees with you. It is also the clearest sign yet that provenance tags like C2PA are turning into live switches, not quiet background data. Below for more…

A work ChatGPT account can feel like your own private desk. OpenAI’s enterprise privacy terms now say otherwise: workspace admins can view, export, and delete employee conversations, the admin rather than the user sets how long chats are kept, and OpenAI reserves more room to hold data past 30 days. Control over enterprise AI chats is moving from the individual who types them to the organization that owns the workspace.

In human terms: You use your company’s ChatGPT workspace to draft and brainstorm. Under the new terms, your workspace admin can open, export, or delete those conversations, and the admin, not you, decides how long they stay. If you assumed your work chats were yours alone, that assumption no longer holds.

Why this matters: This is part of a steady move from “the user controls their own AI history” to “the organization governs it.” That has real consequences for privacy expectations, for compliance and legal holds, and for anyone who treated a work chatbot like a private scratchpad. The direction, not just this one line, is the story. Keep in mind that nothing in OpenAI’s policies restricts an employer from analyzing, parsing, or even fine-tuning their own internal models on employee conversations.

Read below for the details…

Few words in tech get walked back as often as “unlimited.” Artlist just added a new “Unlimited Plans” section to both its consumer and business terms. It defines unlimited as credit-free only, not unlimited speed, capacity, or access, and it reserves the right to throttle, cap, or suspend heavy users, all on the same June 1 that Runway begins retiring its own Unlimited plan. Artlist is embracing the label just as a peer drops it, and the fine print, not the pricing page, is what tells you what you are actually buying.

In human terms: You sign up for an Artlist plan sold as “unlimited” generations. The new terms spell out that “unlimited” means you will not spend credits, not that you get unlimited speed, capacity, or access. Artlist can apply daily limits, slow you down, or suspend you if your use looks heavy, automated, or shared. The word on the pricing page and the fine print underneath it are doing different jobs.

Why this matters: “Unlimited” in AI tools is being walked back across the board. Runway is replacing its Unlimited plan with a plan called Max for new subscribers on June 1, 2026, and moving existing Unlimited users over by September. Phone carriers and fitness apps made the same retreat years ago. Artlist is adopting the label just as a peer drops it, and it is writing in the exact limits that make “unlimited” sustainable. The terms, not the marketing, tell you exactly what they promise - read below for how they get there

The lawsuits hanging over AI tools today are about copyright and training data, not patents. Runway widened the promise it makes to enterprise customers to match: its infringement defense now reaches intellectual-property claims of any kind, where it used to cover only patent claims, and that promise sits outside the contract’s liability cap. It is a real upgrade, but a narrow one, because the promise covers claims that Runway’s service infringes someone’s IP, not claims that the work you generate or the material you feed in does.

In human terms: Your studio uses Runway Enterprise on client work. If a third party sues claiming the Runway service itself infringes their rights, Runway agrees to step in, cover the defense, and pay certain costs, and that promise now covers copyright and trademark claims, not just patents. The limits are where it gets real: Runway will not cover a claim that springs from your Inputs, and it makes no promise that your Outputs are clear to use. So the classic AI worry, that a generated clip copies something it should not, mostly stays with you. (Indemnification just means one side agrees to cover certain legal costs for the other.)

Why this matters: Copyright and training-data lawsuits are the live risk in AI right now, not patents, so moving the trigger from patents to all IP is the right direction, and keeping IP indemnity out of the liability cap gives it real teeth. The catch is what the promise reaches. It protects you when Runway’s service is what infringes. It does not protect you when the claim traces back to your Inputs, to an application you built on Runway, or to the legality of an Output, which Runway expressly does not warrant. For a studio weighing risk on client work, the upgrade is welcome, but it is not cover for the output-level copyright exposure that worries people most. Mechanics follow below:

As AI features get expensive to run and software margins thin out, advertising keeps turning up in tools that never carried it. Cascadeur, a 3D character-animation tool, replaced its bare 2019 privacy policy with a full 2025 rewrite that mostly catches up on US state laws and names a data protection officer, but also brings in Google and Meta advertising cookies and lists targeted advertising as a reason to collect your data. An animation tool has grown an advertising layer, and that drift, creative tools edging into the ad business, is the part worth watching.

In human terms: You visit Cascadeur’s site to grab the animation tool. The refreshed policy says the site can now set advertising cookies from Google and Meta and use your information for targeted advertising, if you consent. The 2019 policy this replaced had none of that. For a tool you reach for to make animation, an advertising layer is a new and slightly odd ingredient.

Below for more …