Generative AI Policy
Most Generative AI (GenAI) technologies are prohibited from use in this jam.
Only the use of approved GenAI models is allowed, subject to the strict rules and conditions below. If you use approved GenAI and do not follow all rules listed in this policy, your submission may be removed.
Our requirements are narrow. We prohibit the use of GenAI models which are trained on content without the permission of its creators. This rule prohibits the use of the most popular GenAI products. For example: DALL-E, ChatGPT, Stable Diffusion, Midjourney, and GitHub Copilot are all prohibited.
So then what's allowed?
The exception to our general ban on GenAI is when jammers use approved models and (importantly) disclose the use of them on their submission page.
A model is considered approved if the whole of its training data is any of the following:
- Owned in full by the model creator
- Permissively-licensed (ie. was released with a pre-existing license which already allows use in GenAI).
- Collected via opt-in means with express permission from the creator.
- Opt-out data collection is unacceptable.
If you are having trouble finding information regarding the training data of a model, please note that if a model does not readily disclose the source of its training data on its official website, it is likely banned.
There are currently not many models which pass this test. Some examples which do include Codeium and Sloyd. Notably, there are not many image generation models which pass this test.
It is up to jammers to research and verify whether the GenAI models they would like to use pass this test, and to credit the use of these tools on their submission page. If you are unsure, please ask an Admin on our Discord, or email us at mail@screamzone.org.
We also consider content created using GenAI to be "non-original", and therefore it has aditional restrictions under our Original Content Policy. This includes ensuring that you create some assets for your game without assistance.
Why is this policy in place?
GenAI is an exciting technology with a lot of potential. We recognise that there are many legitimate use cases for GenAI, especially when not explicitly creating new content (eg. proofreading). These technologies are becoming more widespread over time as companies continue to insert it into their existing products and services.
However, the creative use of GenAI is contentious. Most GenAI models are trained on data obtained without the permission of its sources. There is ongoing controversy around the ownership and distribution of products made using gAI models, and its use can contradict with our aim to encourage jammers to learn and try to create their own assets.
We don't believe that banning GenAI will stop jammers from submitting games with generated content. By banning it, they will simply mask the fact that they used it. By creating a sound policy, we are moving forward in good faith and with hope that our jammers can use GenAI in a way which doesn't stifle creativity, doesn't infringe on the rights of others, and fits into the spirit of our game jams.
What can I do if I see a submission using GenAI against the rules?
Please raise concerns with us by messaging an Admin on Discord, or via email at mail@screamzone.org, and we will take a look. Please keep comments on the game’s page constructive, and avoid telling someone explicitly that they are breaking rules unless this is confirmed by an Admin.
Recap
In order to use GenAI, you have to follow this checklist:
- The model you would like to uses training data which is any of the following:
- Owned in full by the model creator
- Permissively-licensed (ie. was released with a pre-existing license which already allows use in GenAI).
- Collected via opt-in means with express permission from the creator. (Opt-out unacceptable)
- You explicitly disclose the use of the model, including its name and a link to it, on your game's submission page.
- Your game is primarily not GenAI content, as described in our Original Content Policy.