Vol 26, No 15 · June 12, 2026
The_Ledger
AI news is noisy, it's difficult to separate reality from marketing hype and click-farming. Especially in the media & entertainment space.
This project attempts to filter the noise by watching the following signals: tracking exactly what companies tell us through their terms of service, and how their outputs look side by side.
The Ledger is just starting, and kinks are getting worked out – your feedback is wanted.
What changed in AI platform agreements this week
We are tracking over 60 companies and 280 policy documents daily - and growing. Things that govern how your data is handled, who owns what and who’s on the hook when something goes wrong. These are things you should be reading, but don't have the time for - plus they are super confusing.
If something truly concerns you, reach out to a lawyer - there are no guarantees we catch all changes, something vital to you may be missed or misinterpreted. Docs are diff'ed and summarized by multiple LLMs, then researched, edited and published by a non-lawyer human.
Anthropic published a new Privacy Policy on June 8, effective July 8, for consumer accounts only (Free, Pro, Max), not Team, Enterprise, or the API. It adds a verification-data category that can include a government ID and facial geometry, rewrites the standard for sharing data with law enforcement, and for the first time describes Claude taking actions inside third-party apps on your behalf. Anthropic's cover email was unusually candid by industry standards, though it left some of the more consequential details one layer deeper in the document itself.
In human terms: A motion designer on a Pro plan reads the email, sees "we don't sell your data, Claude stays ad-free, you control training," and files it as routine. Those three statements hold. What the email does not lead with is that the same revision describes collecting a government ID and facial geometry if asked to verify age, and broadens who Anthropic can hand data to and on whose say-so. None of it is hidden. It is one summary layer above where the substance lives.
Why this matters: Most of this is genuine clarification, but two items are more than that: the biometric-class verification data is new collection, and the law-enforcement disclosure standard is looser than the one it replaced. Freelancers and studio staff who opened consumer accounts on a work email should also note a new clause that can link those accounts to an employer's enterprise relationship. Below for more…
Sometime between early May and early June, Ideogram rewrote large parts of its Terms of Service. The dispute-resolution section lost its named arbitration forum and its cost protections, the governing-law clause is gone, indemnification now reaches ordinary use, and the liability cap went up, all under a "Last revised" line still reading August 14, 2024. We have the earlier version and the live version side by side. They are not the same contract.
In human terms: Say you have been generating campaign art in Ideogram since last year, and a billing dispute comes up. Under the version you remember, you could read the JAMS rules, lean on the cost protection if arbitration got expensive, or take a small claim to court. Under the version actually in force, none of those exist: no named forum to look up, no cost backstop, no governing-law clause. The date on the page told you nothing had moved.
Why this matters: Ideogram sits in real design and marketing pipelines, and every change points the same way: fewer named protections, no stated governing law, broader indemnity, a higher liability ceiling. The new bar on removing watermarks is the line creators should read closely, because it turns provenance preservation into a contractual duty. And a frozen revision date means anyone who checked whether the terms changed would conclude they hadn't. Below for more…
OpenAI's Service Terms gained a section governing software it delivers for installation on a customer's own systems. The new "Licensed Materials" terms cover code, containers, and modules run on local machines or private cloud, and require permanent deletion of every copy on termination. Contract language for on-prem delivery tends to ship before the product it is written for.
In human terms: A business evaluating a locally deployed model for security-sensitive work reads the new section and finds the exit cost spelled out: when the contract ends, every copy of the installed software has to be deleted. That is a planning item, not a footnote, for anyone building a pipeline around on-prem inference.
Why this matters: This is the kind of clause that signals a product direction before any announcement, which is exactly why it is worth catching here. For studios weighing local deployment, the deletion-on-termination obligation is the term to plan around. Below for more…
When ComfyUI's commercial Terms landed in May, our read was that the pixels are yours but the recipe isn't: the no-training pledge protected your Input and Output, while a carve-out reserved "workflow structures, and node configurations" as metadata Comfy could use.
This week we discovered a new Desktop Privacy Policy, effective June 3, which is the other half of that line, and it is genuinely protective. On your own machine, even with telemetry on, your workflows, prompts, outputs, and model weights never leave the device, which is what makes the Cloud boundary stark: the recipe stays yours right up until you sign in and run it on the Comfy Products, where workflow structure is the metadata the Terms reserve.
In human terms: A VFX freelancer spent two years on her workstation building texture-synthesis workflows for ad clients: the node sequence she refined by trial and error, the sampler choices, the ControlNet chained to her in-painting pass. That craft lives in JSON files on her drive, and the Desktop policy keeps it there, since even with telemetry on the workflow content is not transmitted. Then a brief requires the job to run on hosted infrastructure, so she clicks "Launch Cloud." The workflows transfer cleanly, and her workflow structure, the thing she actually got paid for understanding, is now metadata Comfy is permitted to use to improve its products. The pixels she generates are protected. The recipe that produced them is not.
Why this matters: ComfyUI sits underneath a large share of professional VFX, ad-studio, and game-studio pipelines, which is why the boundary is worth tracing exactly. Outputs commoditize and prompts are noisy, while a workflow encodes expertise: which nodes chain to which, which sampler an artist preferred, an order of operations worked out over months. The Desktop policy confirms that expertise stays local on your own machine. The Cloud carve-out is where it becomes product-improvement input, with no cap, no anonymization requirement, and no time window, while the no-training pledge covers only Input and Output. The contract never says Comfy is building a workflow-savvy agent on user behavior. It does reserve the ability to do so… Below for more…
Forma Construction Connect now includes one integration instead of two for orders "purchased or renewed on or after June 7, 2026," with more "available for an additional fee." Because it applies at renewal, existing customers lose the second connection when they renew, not only new buyers.
The customer-data protection clause dropped its "Cloud and Hybrid Deployments" qualifier, so the safeguards commitment and DPA coverage now read as deployment-agnostic. Small, and in the user's favor.
"Zaps" became "Zap workflows" across both documents. Terminology, not policy.
Added a routing paragraph for exercising data rights through in-account controls, privacy.openai.com, or dsar@openai.com. Process detail, no change to the rights.
Misc Scraper Notes ElevenLabs' 11ai Terms returned a 404 this week; we are checking whether the terms were withdrawn or the URL moved. Perplexity's Third-Party Models page has been behind a bot challenge for two straight weeks. Udio, Cascadeur, and Griptape each failed to capture, for reasons on our side or theirs. The terms are still live; we just couldn't read them cleanly this week.
One prompt. Multiple platforms. Let's give notes.
Bringing decades of experience in feature film, episodic, and commercial production, we evaluate every output against what matters: does it meet the bar for professional use? Think of it as a dailies review for one-shot prompts.
Prompt Tested
Create a photorealistic 5-second video from the input image. The camera remains completely locked off with no camera movement, zoom, reframing, or cuts. The woman maintains direct eye contact with the camera throughout the entire shot. Action sequence: The woman naturally raises the cigarette already in her hand to her lips. She places the cigarette against her mouth and takes a realistic inhale. The cigarette ember briefly glows brighter during the inhale. After a brief pause, she lowers the cigarette away from her mouth while continuing to maintain eye contact with the camera. She then slowly exhales a visible plume of smoke from her mouth toward the camera. The smoke moves naturally and realistically through the air. The shot ends with the woman still looking directly into the camera. Requirements: Preserve the exact identity of the woman from the source image. Preserve facial features, skin texture, makeup, hair, clothing, jewelry, and background. Maintain realistic human motion and anatomy. Maintain direct eye contact with the camera for the entire duration. Preserve the original framing and composition. Realistic cigarette handling and finger placement. Realistic inhalation and exhalation behavior. Realistic smoke density, turbulence, and dissipation. Consistent lighting and color throughout. No facial warping. No identity drift. No hand or finger deformation. No blinking artifacts. No background changes. No additional objects appearing. No scene transitions. No camera movement. High-end cinematic realism.
Realistic smoke and fluid simulations are notoriously difficult to create and render - entire VFX companies have been built upon an ability to do so. We wanted to put this front and center in this test.
LTX 2.3
Veo3.1
Seedance2.0
Wan2.2
Sora2
Kling2.1pro
It seems we've stumped the machines here - none is an excellent result. Kling and Veo may have faired the best for smoke: realistic motion and density but they really are not passable for professional use. This seems like something to either shoot for real, or use traditional VFX, 3d/2d elements to simulate.
