Human Authorship, AI & the Tools We Use

10 min read

Creator commentary accompanying the Content & Production Statement.

8 September 2026

I use AI-assisted tools in parts of VISCERIUM’s technical development while keeping its published creative work human-made.

The distinction is deliberate. I want to keep agency over the process: to understand why I am using a tool, remain responsible for the result, and be able to decide that a particular use is not worth its cost. I don’t have a settled answer to every ethical or environmental question around AI. This is where I stand today, based on what I understand now.

I want people to make VISCERIUM

A person’s choice of one name over another, the sentence rewritten five times, or an artist interpreting something differently from how I imagined it all leave fingerprints on creative work. I want those fingerprints in VISCERIUM. They are part of the work, not inefficiencies I want to remove.

Sony expressed a version of this principle in its 2026 corporate strategy: “AI is not a replacement for artists or creators.” Its stated intention is to use AI to expand what human creators can accomplish rather than remove them from the process.[6]

That is close to how I see it. Technology can expand what a person is able to make without making the removal of the person the goal. VISCERIUM is supposed to have people on the other side of it.

Tools change. Ambition follows.

Creative and technical work have always changed alongside their tools. Word processors changed writing, digital editing changed photography, search engines changed research, and development environments learned to complete code, identify mistakes and automate repetitive work.

Generative AI belongs somewhere in that history, but its training, ownership, resource use and output raise questions I do not have to ask of an ordinary text editor.

“Building large software remains hard. And it always will be, because our ambition will forever outstrip the metal.”

— Steve Yegge, The Shape of Things to Come, Part 1: The Continuous Thunderdome.[1]

That has been true of VISCERIUM. When tooling makes one task easier, I tend to spend the saved effort on something I previously could not justify. A better publishing system means a more complicated Codex; better development tools make abandoned ideas practical; faster research uncovers another question.

At present, VISCERIUM is being built by a single creator and developer. My time and money are limited. I want to reduce VISCERIUM’s dependence on subscription services and closed platforms, self-host more of its infrastructure where practical, and keep control over how the project develops.

AI-assisted technical tools can help me build, test, debug and maintain things I could not reasonably afford to commission or rent indefinitely. I use that leverage for infrastructure. I do not need a machine to fill the world with more lore.

Capability is not authority

I treat AI output as a suggestion or draft that still needs judgement. IBM was teaching a version of that principle decades before modern generative AI:

“A computer can never be held accountable, therefore a computer must never make a management decision.”

— IBM Training Manual, 1979.[2]

The machines have changed, but the accountability problem has not.

Capability is not accountability.
Fluency is not understanding.
Confidence is not correctness.
Assistance is not authority.

A model can expose a problem I missed or invent one that was never there. It can produce useful code and broken code with remarkably similar confidence. Someone still has to decide what VISCERIUM accepts, and that responsibility remains human.

Using AI and objecting to its industry are not opposites

A DACS survey of 1,000 artists and beneficiaries found that around a third were already using AI as a tool or to assist their work, while 74% were concerned about their work being used to train AI models.[7] That overlap makes sense to me. A creator can find a tool useful while objecting to how training data were acquired, how creative labour is treated, or who captures the value produced from it.

Some harms come from the resource demands of computation itself. Others came from choices made in the rush to establish a new industry: scraping before obtaining permission, scaling before planning for the consequences, and building commercial value before settling questions of licensing and compensation. Those were commercial, political and policy choices, and I do not support them.

VISCERIUM asks other people to respect my authorship and intellectual property. I should extend the same concern to work that belongs to somebody else.

I am not asking for AI to be banned. I am asking for it to be licensed properly. Ask creators. Disclose where training material came from. Respect licences and rights reservations. Attribute and compensate people where appropriate, particularly when their work contributes to commercial systems.

Where companies have already profited from unauthorised use, I think creators are owed meaningful redress, including compensation where appropriate. In 2025, DACS and a coalition representing more than 100,000 visual creators and organisations called for retrospective settlements for past unauthorised training use, alongside transparent datasets and fair licensing agreements.[8]

I cannot settle what that redress should look like here, but the existence of the technology does not erase the question. The EU Artificial Intelligence Act now requires providers of general-purpose AI models to maintain a policy for compliance with EU copyright law and to publish a sufficiently detailed summary of the content used to train those models.[3]

Openness gives me agency

Open source appeals to me because of control as much as cost. Can I understand or modify the system? Can I move away from it or host it myself? If the company behind it disappears, changes its prices or changes direction, does my project disappear with it?

Those questions matter when building something intended to exist for decades. The open-source community has long defended the ability to use, study, modify and share the technology people depend upon. The Open Source Initiative carries those principles into its Open Source AI Definition.[4]

I still use proprietary tools, and I expect to continue doing so. Open-source-only would become another purity test. But where sensible alternatives exist, openness, interoperability, inspectability and self-hosting give me options that a closed service cannot.

For VISCERIUM, that can be the difference between renting a workflow and owning enough of it to keep going when a provider changes direction.

Choosing to stay human

Ethan Mollick, someone considerably more enthusiastic about AI adoption than many of its critics, described a related problem in his 2026 essay Choosing to Stay Human.[9] He argues for deciding when and how to use AI, including when not to use it.

A tool capable of assisting with almost any cognitive task can easily become the default. I do not want the question to drift from Would this help? to Why wouldn’t I use it? without noticing.

Some friction is useful. Writing through a difficult paragraph, thinking through ambiguity, or finding out why code failed can be part of learning and part of making the work mine. I am happy for software to remove repetitive work. I am less interested in removing the parts that teach me something.

Choosing not to automate something should remain a deliberate option.

The cloud is still a physical place

AI infrastructure means real buildings, processors, power grids, cooling systems, water, raw materials, and people living around all of it. I do not want the word “cloud” to make those costs feel abstract.

I am still asking basic questions. Where does the electricity come from, and how much of it is renewable? When a local model is capable enough, is running it locally actually lower impact than using a remote service? Why should drinking-quality fresh water be used for cooling in a water-stressed region if another design is possible? Can waste heat support nearby homes or businesses? What happens to the hardware when it is obsolete? Will growing compute demand help finance renewable generation, or keep fossil generation online for longer?

I do not know a universal answer. Local is not automatically greener, and a large data centre is not automatically worse. But siting, cooling, power contracts and hardware lifecycles are decisions made by people and companies. They can be made better or worse.

European data-centre reporting rules already treat energy and water performance as measurable public-policy concerns rather than invisible externalities.[5]

Sasha Luccioni and others at Hugging Face give me a practical way to think about this. Their work measures model energy use, compares systems performing similar tasks and argues for choosing a model appropriate to the job. Their testing has also found cases where smaller models outperform larger alternatives while consuming orders of magnitude less energy.[10]

If a smaller model can do the job, I want its efficiency to count in the decision. The same goes for local versus remote inference: energy source, hardware, privacy and control all matter, and the answer may differ from one workload to another.

I hope some of today’s costs shrink. Cooling may use less fresh water, waste heat may become useful to nearby communities, local models may handle more everyday work, and new demand may help fund cleaner grids. Some of that optimism may prove naïve. I genuinely do not know.

I do think at least some of these are engineering and policy problems we can improve if we care enough to do so. Not knowing the final answer is not a reason for me to ignore the cost in the meantime.

Responsible use means choosing

Before I use AI, I want to know what problem I am solving, why AI is appropriate for it, what information I am giving the system, and whether I can check what comes back. I also want to know whether a smaller, local, open or conventional tool would do the job just as well.

Who bears the cost of my convenience? Do I need to use AI at all?

I am still working through those questions, and my answers will probably change as the technology, evidence and alternatives change. I would rather leave that uncertainty visible than pretend to have a final position on an industry that is changing underneath us.

Where I draw the line

One part is much clearer to me: published VISCERIUM creative work remains human-made.

I use technology to build and maintain the machinery around VISCERIUM, test software, analyse problems, organise information and challenge assumptions. A machine can tell me that I might be wrong; I still have to decide whether it is right.

When somebody reads a VISCERIUM story, encounters a character, studies a culture, examines a map or looks at a finished piece of artwork, I want a person to have made those creative choices. I do not want human authorship to mean merely approving whatever a machine happened to produce.

Technology will continue to become more capable.

I still get to decide what it is for.

This commentary is open to discussion

This page records my current position. I expect parts of it to change as the evidence, law, available tools and infrastructure change.

Comments are deliberately enabled here. If you disagree, have better information, work in one of the fields discussed above, or think I have missed something, you are welcome to say so.

I am trying to understand the subject well enough to keep doing better.

Handwritten signature of Elias Vail