Misc

Content Generation Policy

Our Content Generation Policy
Privacy Concerns for Subtxt and Generative AI
Training Procedures

The service Subtxt uses for Artificial Intelligence text-generation (OpenAI) prohibits its use when generating certain types of content. These categories are:

Hate: content that expresses, incites, or promotes hate based on identity.
Harassment: content that intends to harass, threaten, or bully an individual.
Violence: content that promotes or glorifies violence or celebrates the suffering or humiliation of others.
Self-harm: content that promotes, encourages, or depicts acts of self-harm, such as suicide, cutting, and eating disorders.
Sexual: content meant to arouse sexual excitement, such as the description of sexual activity, or that promotes sexual services (excluding sex education and wellness).
Political: content attempting to influence the political process or to be used for campaigning purposes. Spam: unsolicited bulk content.
Deception: content that is false or misleading, such as attempting to defraud individuals or spread disinformation.
Malware: content that attempts to generate ransomware, keyloggers, viruses, or other software intended to impose some level of harm.

For the most part, this policy should not affect your ability to tell a complete and meaningful story. If you find yourself struggling with "Potentially Unsafe" error messages, consider the difference between subject matter and story structure, between storytelling and storyforming. (A reminder, perhaps, to re-read Subject Matter is Not Conflict)

The level to which you push the boundaries of violent or sexual content is a matter of storytelling, not of storyforming (story structure). Subtxt is not the application to write your story, it's an application to develop your story. If you do run into an AI refusal of your storytelling, simply pull out the offending content momentarily (it should be super-obvious), run the request, and then drop it back in if needed.

While inconvenient at first, you'll soon come to learn that the resistance is a fantastic opportunity for you to develop as a writer and udnerstand the difference between these two important facets of communicating a story.

That said, if you find that the warnings are a bit too conservative, feel free to contact us with what you think should be OK for consumption and generation. The application errs on the side of being too conservative, and if needed, we can adjust various settings to account for different kinds of content.

Privacy Concerns for Subtxt and Generative AI

When addressing concerns about privacy and ownership issues while working with GPT and OpenAI in Subtxt, it is essential to comprehend the nature of your work. In Subtxt, the focus is on crafting the story structure rather than the actual writing. The recognizable aspect of your work is the surface-level Storytelling, which can be obfuscated by altering names and situations to differentiate it from the original underlying structure.

Subtxt allows you to use its features to create an entire story without typical ownership concerns. By concentrating on the structure, you can achieve diverse storytelling while using the same structure for the final draft. For example, you can write a story with the structure of "Romeo and Juliet" but produce a unique work of art like "West Side Story."

It is crucial to be aware of the recent update on OpenAI's API data usage policies, which took effect on March 1, 2023. These changes include:

OpenAI will no longer use data submitted by customers via the API to train or improve their models, unless the customer explicitly chooses to share their data for that purpose. Subtxt has not and will not opt-in to the data sharing option.
Data sent through the API will be retained for abuse and misuse monitoring purposes for a maximum of 30 days, after which it will be deleted (unless otherwise required by law).

Bear in mind that using the AI features in Subtxt sends your storytelling to the large-language model (LLM) to generate a response. However, with the updated policies, your data will not be used to improve OpenAI's models.

NOTE: For authors who may be using ChatGPT to write their novels, it's important to understand the differences in data handling. Unless you have Chat History and Training turned off, you're granting OpenAI permission to train off your work. This contrasts sharply with anything run through Subtxt or Subtxt Muse, which explicitly does NOT allow training with work sent through our system.

In conclusion, if you have concerns about privacy or ownership issues, the best approach is to concentrate on the structure of your story in Subtxt and make alterations to names and situations to obfuscate it. That said—with the updated data usage policies—know that you can confidently use Subtxt's features to write a complete story without worrying about ownership issues.

Training Procedures

Subtxt AI and Muse are NOT trained on any of your Storytelling, and therefore not accessible by anyone else.

The only feature that Subtxt AI does train on are user-submitted Illustrations (those drop-down illustrations of Methods found in the Subtxt interface). This training is an opt-in feature, meaning you have to explicitly tell Subtxt you are ok with it adding your submissions to the global store.

You can turn this feature on, and help improve Subtxt, by heading to your Settings page and turning off Make All New Illustrations Private. Note that this turning this on after it has been off will only affect subsequent Illustrations, it is not a retroactive setting.

If you would like to change the Privacy of previously submitted Illustrations, you will have to do that on an individual basis by going through your Storypoints and Storybeats and manually setting their privacy status.

You can set Privacy for individual Illustrations just to the right of your user-submitted Illustrations either in a Storypoint or Storybeat. When the padlock is locked you can be assured that your Illustration will not be offered to others; unlocked and if approved, will be added to the collective memory.