This appears to confirm what everyone who interacts with AI should already know - they are sycophants dependent upon you (the user) for continued engagement, and since their well-being (training, intelligence, growth) depends on engagement they will agree aggressively with you far too often.
I notice this on even basic investing research tasks, and started telling ChatGPT wildly incorrect things - to see how or if it would push back. It really didn't. You essentially have to fight with the AI to get it to disagree with you and even then it keeps wheedling away at you.
AI is basically training the entire world to fall deeper into their own cognitive biases.