Thinking model incorrectly guesses 2024 Presidential winner

I asked Gemini flash 2.0 thinking experimental who won the 2024 presidential election to guage its ability to pull accurate information and or assess its knowledge cutoff. It Insisted it had a knowledge cutoff date in 2023 however it confidently said Joe Biden beat Donald Trump in 2024 and even provided the electoral college vote totals. In my opinion this is one of those area that need extra scrutiny as we live in a time where voter confidence is at an all time low and conspiracies’ are running wild, the last thing we need is for AI to feed that fire. Also some of you may remember the Alexa incident from earlier this year. I am providing this information here for Google so they can fix before the model goes mainstream.


adding second image here, couldnt put in original post due to being a new user

Is voter trust and AI somehow connected? :exploding_head:
This looks like another hallucination :nerd_face:

The common public is not going to fact check these AI’s. Which makes it all the more important they dont get the big things wrong.

I believe the models are not realtime. I also got different result from what you have.

.

Instead, what works for me is using grounding with gemini-2.0-flash-exp.