AI hallucination
When an AI makes up a fact, quote, or law and states it with complete confidence.
What is an AI hallucination, and how do I avoid getting false information from it?
Language models write the most likely continuation of a piece of text. That almost always matches reality, but sometimes the result sounds perfect and is completely made up: a number nobody ever published, a law cited with the wrong section number, or a code library with a name that just resembles a real one.
It is called a hallucination because the AI never flags it. It writes true and false statements with the exact same confidence, so anything that matters needs to be checked outside the AI, against the original source or by actually running the code.
Why it matters to your business
If you use AI to write your website copy, answer customers, or prepare a report, a hallucination can end up published under your name, like a price you do not actually offer or a regulation cited with the wrong number. That is why it helps to have the AI work from your business's own information, with a person reviewing anything before it goes public.
How we use this at Damorus
On the Damorus site we follow a written rule: no figure gets published without a source, and every fact in this glossary about laws, dates, or payments was opened and read at its source before we wrote it down. La Roca taught us another lesson that applies equally to people and machines: the absence of a visible error does not confirm that something went right. When we could not verify something, we said so.
Need this on your project?
I want my free mockup →