Are your AI video models compliant?

A few weeks ago, we were getting ready to ship our storyboard to video workflow. The whole point of it is character consistency: you define your actors once, and they stay themselves across every scene. To do that, we send reference images of those actors into the video model on every generation.

And every one of those requests started coming back rejected.

The model we were using for high-quality video at the time was Seedance, from the Chinese company ByteDance (the same company that owns TikTok). It turns out Seedance runs a face-detection pass on every uploaded image before it even reads your prompt, and it hard-blocks any image with a recognizable human face. Our actors were the whole feature, and the model was refusing to look at them. This is well documented: creators hit the same wall constantly, and even third-party API providers warn that "images with real human faces are blocked by ByteDance content policy." No prompt rewrite fixes it, because the prompt is never the problem: the image is rejected at a separate scan stage before the prompt is read.

We could have shipped a workaround. Instead we moved the whole tier off Seedance. And the timing turned out to matter a lot more than we expected.

This was never just our problem

While we were dealing with a content filter, Washington was busy turning "which AI model are you using" into a compliance question.

The U.S. government is tightening two related but distinct sets of rules around AI.

The first is already law. The Department of Justice Data Security Program restricts certain transactions that give China and other "countries of concern" access to bulk sensitive U.S. personal data or government-related data. It was issued under Executive Order 14117, took effect on April 8, 2025, and moved into active enforcement on July 8, 2025. China, including Hong Kong and Macau, is one of the six designated countries of concern. Covered categories include biometric identifiers, precise geolocation, personal health data, personal financial data, genomic data, and certain personal identifiers. It is not a blanket ban on sending any data outside the United States: applicability depends on the type and volume of data, the transaction, the provider, and the provider's ownership or jurisdiction. See also the Federal Register final rule.

The second set of rules is newer and moving fast. Just this week, Treasury Secretary Scott Bessent threatened sanctions against Chinese AI companies over intellectual property theft, and the White House accused Moonshot AI of distilling Anthropic's Fable model, with Entity List designations "on the table." Reporting from Tom's Hardware and MLQ describes a broader push (procurement restrictions, security advisories, and liability rules) that stops short of a formal ban but creates real friction for any U.S. business built on a Chinese model.

It is not only about one model, or one country

The lesson we took away was not "Seedance bad." It was that every model you build on comes with a policy, a jurisdiction, and an availability window that you do not control, and any one of them can change under you overnight.

• The Chinese video models are the most obviously exposed. ByteDance's Seedance, Kuaishou's Kling, and MiniMax's Hailuo each combine aggressive, opaque content filters with a regulatory target on their backs.
• Moonshot AI's Kimi, billed by its maker as the largest open-weight model in the world and popular enough that Microsoft has reportedly been testing it for Copilot and Azure, is now at the center of the sanctions fight. Teams that standardized on it may suddenly need an exit plan.
• Even Western providers carry surprises. ElevenLabs, a leading (non-Chinese) voice company, quietly bars government entities from using its service without prior written authorization in its Prohibited Use Policy. If you sell into the public sector, that clause can block you just as hard as a geopolitical one.

None of this is about picking the "good" providers once and moving on. The set of acceptable models is a moving target, and it moves for reasons that have nothing to do with your product.

What VideoGen does about it

We do not think you should have to track any of this to make a video.

VideoGen is the model layer for AI video. Our model router sits between your workflow and the underlying providers, and it decides which model runs each generation based on quality, availability, and eligibility, not on which logo happens to be trending.

Concretely, after the Seedance episode we moved high-quality video generation onto Google's Omni model, with automatic fallback to other vetted, non-Chinese providers. If a model rejects a request, goes down, or becomes unsuitable, the router routes around it on that request, transparently, without you rewriting anything. Our actors generate cleanly now, and the compliance posture of the stack is something we manage so you do not have to.

We continuously evaluate the models and providers available through VideoGen, including:

• Where customer data is processed
• Which company operates each model
• Whether data is retained or used for training
• Which models can be used for particular customers or regions
• How to replace a restricted model without breaking the workflow

This is not a legal guarantee that your company is compliant. Compliance depends on your data, your customers, and your counsel. What we can substantiate operationally is model eligibility, data-routing controls, and fallbacks that keep your workflow running when a model becomes unsuitable.

The takeaway

If you are building AI video into a product or internal workflow, the single most valuable thing you can do is stop hard-wiring your stack to a specific model. Inventory what your stack actually calls (including through aggregators), ask where inputs are processed and retained, and prefer a platform that can swap models without you shipping a code change.

That is the entire premise of VideoGen. Build on the model layer, and you never have to wonder whether today's model is the compliant one, because that is our job, not yours.

Start creating with VideoGen, or talk to us about model eligibility for enterprise or regulated use.