When AI gets it wrong, and how to check

AI can invent facts, numbers and quotes that sound right. Ask for sources, check the key figures, and never publish what you have not read.

The labs are improving: OpenAI reports that GPT-6.1 Terra's error rate on its hard test prompts fell from 53.6% to 16.7% at low reasoning effort. That is better, not perfect.

A three-step check:

  1. Ask for sources and open at least the ones behind your key claims.
  2. Re-derive the numbers that drive a decision, or ask a second model to check the first.
  3. Ask it to argue the other side: "What would a sceptical CFO say about this?"

Be extra careful with law, tax, medical and anything about real people. Treat AI output as a strong first draft from a capable colleague who has not been to your office.

Sources

Learn it hands on, with us.

Free AI build nights in Kuala Lumpur. Bring one real problem and leave with a working tool.

Sign up for next gathering