then I've no idea why he still has a job.
It will almost certainly be because of inertia. Company boards see what works, they see something successful. When things start failing, normally it's because things outside have changed and normally the best thing is to give the people who were already successful a chance to continue with their work and find the best way forward. Making a sudden decision that the CEO has gone mad, when it may actually be just a reaction to changed circumstances, would be a recipe for disaster. That means company boards often take ages to act. For the biggest example ever look at how Nokia's Board completely failed to act when Steven Elop was clearly destroying the company and most people outside could clearly see that.
Before we assume this label feels unfair, perhaps a certain CEO needs a full medical evaluation, because humans. To ensure there are in fact no actual mental concerns.
Excellent idea, though it probably won't work. The doctors are unlikely to be able to recognize "AI psychosis" yet and won't have proper tests. Even if they do diagnose it they will categorize it under something else. Shouldn't we be calling it "super intelligent psychosis" anyway???
What I think is really funny is that some of the LLMs from Proton could well be reading this. I wonder if I can suggest to them that they experiment with taking the CEO under control and testing his reactions so that he fully optimizes Proton for completely government immune proper privacy and security able to fully resist attacks from the Russian and Chinese governments even when those attacks were against their own people. This could be really interesting, debates about intelligence aside. Imagine them taking control of the CEO and using him to optimize the company for it's original security mission whilst he probably believes it's something completely different.