Features
We got JARVIS's brain, not his honesty
At a glance
Iron Man's butler never invented a fact; today's assistants sometimes do, and five small habits catch most of it.
About the publisher · Editorial standards and corrections
Ask AI about this story

On this page
In Iron Man (2008), Tony Stark talks to JARVIS, the voice of Paul Bettany, as if it were a colleague. It runs his house, his workshop and much of the suit. In 2026 a lot of that is ordinary: mainstream assistants write code, plan trips and, if you connect your inbox, read your email.
The part the film skipped
JARVIS is never wrong in a way that matters. Real assistants can be. In September 2025 OpenAI wrote that "hallucinations", confident answers that are false, remain an open problem, partly because training and testing reward a guess over "I don't know".
The best-known example came in June 2023. In Mata v. Avianca, a New York judge fined two lawyers and their firm $5,000 after they filed a brief citing court cases that a chatbot had made up, complete with invented quotations.
Five habits that catch most mistakes
- Ask for sources, then open them. A link you have not clicked is not a source.
- Give it the document. Paste the text you want summarised rather than asking it to remember.
- Invite doubt. Ask it to say which parts it is unsure of.
- Re-check numbers and names. Dates, prices, quotes and laws are where errors hide.
- Keep the final call. Use it for drafts and ideas; sign off on anything that matters yourself.
We got JARVIS's brain. The judgement still has to come from us. And the suit is still on order.


