As a customer, especially corporate, the PR shitstorm is definitely connected regardless of whether it's just a system prompt that was changed on the X version. Now that the owner of both X and xAI has demonstrated his inclination to introduce his bias on the main public version of their model, I wouldn't be so confident that training of the base model itself won't be biased too in the future, or perhaps already has been.