How to get factual answers when misleading the AI?

I am able to generate untrue facts by misleading the AI, even when telling it a fact in the prompt. Please see the example below. Albert Einstein did not have a brother but the AI thinks he did if I simply ask about him. The AI seems to be making up information. Changing the temperature did not fix the issue. I’m still able to get around factual answers.

How can I get factual information about factual things like historical figures, even when misleading the AI?

Another example:

Unfortunately, truthfulness isn’t as simple as moving a slider.

OpenAI has done a good job describing how to fine-tune a Classifier to improve the number of factual answers. See the help article here.