Stay Critical!

• Verify AI-Generated Knowledge: Even texts or answers that seem coherent at first glance may contain errors ("hallucinations") (see How Do AI Tools Like ChatGPT Work?).
• AI Bias: AI systems are trained on existing texts. The languages, cultures, worldviews, stereotypes, etc., most prevalent in the training documents are likely reflected in the AI's responses. Other perspectives may be overlooked.
• Source Citation Issues: Always verify the sources cited by ChatGPT.

Take Responsibility

• Mark content created with the help of AI.
  If using AI in academic assignments, discuss it with the person grading the work! In theses, the use of AI tools might need to be mentioned in the declaration of originality. Clarify which AI tools are permitted and how they should be acknowledged.
• If you use incorrect information generated by AI, you are responsible, not the AI.
• Use AI as a Tool. Learn to use AI effectively and select the right tools. (A spoon can dig holes, but there are more suitable tools; sometimes, hands are even faster...)
• Let AI assist with learning rather than doing the thinking for you.

Copyright Aspects

• AI-Generated Content is not protected by copyright.
• The AI's training data may have been copyrighted, and the AI does not reliably cite verbatim, content, or structural quotes. Texts copied from AI might show matches with original texts in plagiarism checks.
• Uploading copyrighted documents (e.g., full-text PDFs of journal articles) into AI tools can be problematic legally, especially if the copyright holder (or licensee) has restricted such use for AI training or text and data mining.

Data Protection Aspects

• Inputs and uploaded content may be used for further AI training.
• Tools like ChatGPT collect personal data such as contact information and message content.
• Data from social media sites and usage (e.g., log data, cookies, usage data) are also collected.

This list of AI tools is not exhaustive and does not reflect a definitive quality assessment. These tools are not centrally provided by THA. Users must individually verify compliance with policies, data protection, and other relevant criteria before use.