AI Policy

AI is increasingly ubiquitous in every field of work. Temerity's approach to AI is guided by our ethics, our responsibilities to our backers, and our personal experience.

Our values regarding the responsible use of AI on the SEAR project

We value artists and their art. We seek to properly compensate and credit artists, and to appropriately license their work.
We value and prefer human creativity and collaboration. It is the reason most of us got into this field.
Should we be lucky enough to grow, we are committed to maintaining space within the studio for entry-level developers, admins, and artists regardless of AI’s capabilities. This is a question of supporting the industry as well as its people by fostering and recognizing the human joy inherent to the gaming industry. Being able to maintain policies like this is why we want to maintain our independence and control.
We believe AI output can be detrimental to a project without a human or human team with the appropriate skills directing its use and refining its output. (AI may be capable of producing generic proto-draft concept art. It is not a concept artist. Nor, in our view, does a person with no artistic skill become a concept artist by virtue of utilizing an AI prompt.)
As a matter of opinion, we believe that the training initially performed for most modern AI tools should have been conducted with the approval and buy-in of artists whose work was used in that training, and in fact, some of us have work which was scraped for that training.
The increasing use of AI in the art fields means it is insufficient to say we will never use AI art and trust that our own judgment and efforts will be sufficient to keep that promise. We need to be prepared for realizing that art we thought was human-made was not.
We recognize AI tools’ growing ubiquity, utility, and capacity to accelerate workflows, and realize there is a point at which it is strictly good business practice to adopt some of those tools when we feel we can do so ethically.
We have an obligation to our backers, supporters, and staff to be open to workflows that will drastically accelerate production and make maximum use of every dollar. All other things being equal, if a six-hour task can be completed in ten minutes, it should be. In aggregate, such decisions can be the difference between completion and failure of the project, and we owe it to our backers to complete the project and keep our promises.
These values will inform every decision we make around how, when, and whether to use an AI tool. We will make mistakes. We will discover issues we were unaware of. But we will never blindly embrace a new tool unless we have reviewed these values and determined the use of the tool is in line with them.

Ways we will never knowingly use AI in the SEAR project

We never use AI-generated music or voices. Doing so would be absolutely antithetical to our goals for SEAR. Our goal is to showcase and celebrate human-created music, especially great music seeking a wider audience.
We never use AI-generated assets as hero art, player pawns, UI elements, fonts, or other prominently-featured art elements which serve as the face of the game or as a primary point of interaction.
We never use AI to draft public-facing communication (such as project updates or community outreach).
We never use AI to draft written in-game communication (such as dialog scripts, hover text, or tutorial text).
We never treat AI output of any kind as a final product.
When we create art as homage, we will not use AI to create that art.

Ways we routinely use AI in the SEAR project

Web searches / LLM responses for research. When done with appropriate skepticism and follow up, this is substantially faster and often more accurate than a straight text search, and is increasingly unavoidable in any case.

Ways we may sometimes use AI in the SEAR project

Fast iteration of concept art and rapid visualization of ideas, always done at a human concept artist’s direction and discretion. AI is not a concept artist. If one of these visualizations forms the seed of a direction we want to go, we switch to an iterative human process.
Placeholder art assets in-game, only in lieu of time-consuming in-house development, always with the intent to replace (see detailed policy below), and only for assets that are strictly background or display only briefly. We prioritize the purchase of available assets online that meet our needs when available and will perform basic due diligence to avoid purchasing assets that were AI-produced. Why would we ever do this? This is especially useful in interim situations where we have a number of different directions a level or effect could go and would like to get a high-level idea of how different aesthetics might present, but it is also useful in situations where we have a highly specific need, there are no such pre-made assets available, and we lack the funds or time to commit to a full in-house development process. Again, this would never apply on a hero/primary asset. If we lack funds or time to commit to a full in-house development process of a primary asset, we will either forgo that asset or use a human-created placeholder.
We may use AI coding tools for back-end development, particularly for rapid iteration and concept testing. Always at the explicit direction and discretion of a knowledgeable human developer or developer team and to be used in the project only after thorough human verification and testing.

How we follow through on replacement of placeholder AI assets

We maintain a publicly-viewable list of every placeholder AI asset currently in the game, whether in a playable publicly-released version or in our development environment. Every asset will be identified on this list by a picture, an in-engine identifier (such as “SM_AIMesh”), a type descriptor (mesh, video, image), a description of its role in the game, and a rationale for why it is in use over a human-authored asset. We will regularly review this list in the context of our revenue and expenses, and whenever a decision is to be made between increasing profit and replacing the assets on this list, replacing the assets will take priority and we will invest as necessary (via contracting, asset licensing, hiring, etc.) to make that happen.

The list will also serve as a public listing of assets we used in the game under the belief that they were human-made, and then later learn were AI-made.

Additionally we maintain a policy of solicitation and openness toward submissions from independent artists who have the skills to replace such assets with human-made versions. Revenue permitting, we will negotiate a fair license fee with any artist who can replace one of our AI placeholder assets with an asset of equal or higher quality that meets the same aesthetic and performance requirements, and add their name to the list of SEAR contributors.

AI art assets currently being used in SEAR

To our knowledge there are no AI art assets currently being used in SEAR.