Because OpenAI cannot offer them competitive compensation, anymore. Not when they can court VCs and get hundreds of millions of seed funding to setup their own AI startup.
Also OpenAI is by this point a mature project so lots of people might feel a bit stifled in their freedom and they might be dying to explore new ideas.
3. Other firms are offering huge incentives to OpenAI employees, and engineers are doing what engineers always do: job-hopping to jump a rank or two in the job ladder. From last year, but still relevant IMO: https://techcrunch.com/2023/11/20/openai-ai-talent-poaching-...
4. People are concerned that OpenAI's future is not as bright as it seems given its current (objective) technical dominance, a discussion centering around terms like "moat" and "propriety data" and "open source vs. closed source". Some challenges facing the company include: quite uncertain lawsuits of existential size (https://www.nytimes.com/2023/12/27/business/media/new-york-t...), the fact they're losing $5B/y without a proven path out of that (https://www.datacenterdynamics.com/en/news/openai-training-a...), and, on a meta level, your question itself -- they've lost basically all of their top scientists.
This is based on reported habits on the top AI labs on their strategies, Google's recent bid of spending billions to essentially rehire an old employee, and the general nature of people switching jobs in the tech world for a higher salary
Partially it's weirdness from Open AI transitioning from non profit to profit, partially it's from people who are delusional about the dangers of AI, partly it's great unheard of offers from other companies.
