Using AI assistants sensibly and safely

In short: AI assistants (chatbots) answer questions and write texts but can make mistakes and sometimes invent facts. Do not enter passwords or private data and check important answers.

In more detail: Language models predict likely fitting words. They do not understand like humans and can be convincingly wrong (“hallucination”). Inputs are stored or used for training depending on the provider.

In Depth

• Good for: Ideas, summaries, explanations, drafts, translations.
• Bad for: Reliable facts without checking, legal and health questions, calculations without verification.
• Check: Compare important statements with reliable sources.
• Privacy: Do not enter passwords, health data, trade secrets or other people’s personal data.
• School/training: Follow your institution’s rules. Do not replace your own learning.
• Good questions: State the exact task, an example and the desired form.
• Copyright: Results may imitate sources. Be careful when publishing.

See also: Everyday privacy: basic rules, Searching the internet properly