What AI cannot do.
Know the limits, keep the responsibility.
6 limits.#
Guarantee truth
Language models calculate what is likely. Facts, quotes and sources may be invented.
Take responsibility
Responsibility for a result lies with the person who uses it.
Understand intentions
AI only recognises context, mood and unspoken goals to a limited extent.
Give legal advice
An AI does not replace legal advice. It can give guidance, but no binding information.
Be up to date
Without search, models only know their training data.
Be neutral
Biased training data leads to biased results.
Who is liable?#
As a rule, whoever uses the system is responsible. More on the AI regulation page.
Questions? Get in touch.
Write to me if a topic is missing or you want to know more.
Sources and further reading
Was this page helpful?
Found a mistake? Let me know.
Suggested citation: Russo, K.: "What AI cannot do". sicherheitsdenker.de/en/ki-grenzen/, as of 6 November 2026.