According to our Content Policy you have to have the rights to work with a text in order to use it in our system. That means if you train a LAIKA brain you’re supposed to only use materials you legally have the right to copy. If you use excerpts, copyright is more lenient and allows more flexibility. In the end it’s often your call – you are responsible for the texts you work with, after all.
Every NDA (non-disclosure agreement, a contract outlining how the covered materials can be communicated and to whom) is different. Our Privacy Policy (new) clearly outlines what happens to your data and who is handling it. In general we do not train models on your materials and only store them for the purpose of running the system. E.g. if you want a summary of your text we necessarily have to upload the text to a server that runs a language model to create the summary. We don’t save the text for any purpose other than processing it according to what you want us to do with the text.
We are happy to talk about a custom deployment in your company.
As the Privacy Policy (new) outlines, we collect your data solely so that we can:
We do not train any language models outside the ones you personally request (using the legacy LAIKA functionality) with your data. We do not use your data to train language models that other people than you have access to.