OpenAI's newest AI model is built to admit what it can't do

OpenAI has a new AI model called GPT-6.1 Terra, and the company says its big selling point is honesty. The goal is an AI that tells you when it is unsure or cannot do something, instead of bluffing its way to an answer. Inc. reported the news on October 27, 2026.

What changed

OpenAI describes the model as more forthright. In plain terms, that means it should be more willing to admit the limits of what it knows or can do. Many AI tools have a habit of sounding sure even when they are wrong. A model that says "I'm not sure" is easier to judge.

Who it affects

Anyone who uses AI for work or money questions has a stake here. A confident wrong answer can lead to a bad decision, such as a mistake in a report or a poor call about spending. An honest "I'm not sure" gives you a signal to double-check before you act.

Inc. also points out that people struggle to admit their own limits, not just AI. That is a useful reminder that this is a hard habit to build in anyone.

What is not known yet

• How OpenAI measured the improvement in honesty.
• How often the model still gets things wrong without warning.
• When and how people can use it, and what it costs.

What to do

Treat any AI answer on money, health or business decisions as a starting point. Even with a more honest model, check important facts with a trusted source before acting on them.