AI Ethics
How we think
about what
we build.
Most AI ethics pages are legal documents dressed up as values. This one isn't. These are the actual positions we hold, the lines we won't cross, and the things we're honestly still figuring out. Read it as a conversation, not a policy.
It's AI. We made it. We said it.
We don't use the word "transparency" about this. Transparency implies there's something to hide. There isn't. The work Wrayven produces is AI generated and directed by human creative judgment. That's what it is.
We're not going to wrap that in language designed to make you feel better about it. You either want AI cinematic production or you don't. Period. If you do, we're the right conversation. If you don't, we're not the right fit and that's completely fine.
"The companies pretending their AI work isn't AI are solving for the wrong problem. The work either holds up or it doesn't. Ours does."
What clients do with the work after delivery is their decision and their responsibility. We produce the creative. We are not responsible for how it's used downstream. The one exception: if a client uses Wrayven-produced work to falsely claim it was filmed traditionally, we will not defend that claim. We will disclose. The work being AI generated is not something we will lie about under any circumstance.
Some work we don't take.
These aren't all moral positions. Some are quality arguments. Some are both. All of them are firm.
Industries and categories we don't work with
"AI testimonial and UGC videos are just a lie. Full stop. What you're selling in those cases is reality — and AI shouldn't be selling reality."
What's running under the hood.
We use Higgsfield (which runs on Nano Banana, Seedream, and Kling models), Adobe Firefly, and Claude for prompt development. Reference images are sometimes uploaded to Claude to assist in prompt generation.
Every client signs a written agreement acknowledging these tools are in use before a single generation is produced. Not buried in terms of service. A direct, explicit written acknowledgment. No surprises.
Client briefs, reference materials, and deliverables are treated as confidential. We ask for explicit written permission before featuring any client work in Wrayven's portfolio or marketing. If you say no, we don't use it. If you don't answer, we don't use it. The default is your work stays yours.
This is not a threat.
Photoshop has "do it for you" tools. Nobody argued Photoshop destroyed design. What it did was separate the people who understood design from the people who didn't — faster and more visibly than before.
AI is the same thing at a higher order of magnitude. It amplifies what you already are. A lazy creative with AI is a more productive lazy creative. A truly talented person with AI becomes something the industry hasn't seen before. The threat isn't to talented people. It never has been.
The real risk is for people who were hiding behind production process — who used the complexity and time of traditional production as cover for the absence of a genuine creative point of view. AI removes that cover. That's not a threat. That's a reckoning the industry needed.
"If your AI creatives look bad, your human creatives most likely look bad too. The tool reveals the standard. It doesn't set it."
Wrayven's own hiring practices reflect this belief. We pay fairly. We build small teams on purpose. We hire people whose judgment is the asset — not people who can execute instructions. That distinction matters more with AI than it ever did without it.
Here's what we actually know.
AI generation uses energy. Data centers use energy and water. A significant portion of both currently comes from sources that cause environmental harm. That's true. We're not going to pretend it isn't.
What we can't do is verify the full environmental practices of every tool we use. Higgsfield doesn't publish environmental reports. Adobe has documented renewable energy commitments and hit 70% wind and solar coverage in 2024 on their way to a 100% target. That's meaningful and it influences our tool decisions going forward. But we're not going to make claims we can't verify about the infrastructure running under tools we don't control.
On water specifically: some modern data centers are moving toward closed-loop cooling systems that recirculate rather than evaporate. Some use immersion cooling that requires near-zero water. That's real progress. But the majority of existing data centers still use evaporative cooling that permanently removes water from local ecosystems — even when that water started as recycled. The nuance doesn't erase the impact.
It's also worth saying: AI production eliminates the transportation, travel, crew logistics, location permits, equipment manufacturing, and physical waste of traditional production. Whether that net calculation is positive or negative depends on variables we don't fully control. We're watching it and we're honest about what we don't know.
"We can't fix this ourselves. Nobody using AI can. The fix happens at the infrastructure and regulatory level. What we can do is fund the organizations making that pressure real."
So here's what we're doing about it.
Where Wrayven donates a percentage of annual profits
We're also making tool decisions going forward with environmental commitments as a real factor — not a marketing checkbox. Providers who can't or won't document their practices are held to a higher standard of proof before we add them to our stack.
Does any of this solve the problem? No. The problem gets solved when data centers are required to use closed-loop cooling and when the grids powering them run on renewables. We can't make that happen alone. But that's actually how environmental progress has happened every time it's happened. Not from individual virtue alone. From enough individual commitments creating enough collective pressure that accountability becomes inevitable. We're part of that pressure. Intentionally.
AI in marketing is like money.
It doesn't turn you into
something you're not.
It amplifies what
you already are.
If you're lazy, you'll be more lazy. If you're amazing, you'll be even greater. If you are a liar, you will lie at a greater level. It's a tool. And a tool's usage depends entirely on the user — not how it was intended to be used.