I think you're missing the point: companies like Google set up research units like this to act like tenured positions, like mini-universities, and then employ a

I think you're missing the point: companies like Google set up research units like this to act like tenured positions, like mini-universities, and then employ academics like Gebru just so they get the credibility that comes with employing them. "See, we're not all about profit, we support basic research in this area, even when it criticizes our other work." Critical to that credibility is actually having the same standards as Harvard with respect to allowing the researchers to publish what they will, whether or not they're the target. By appearing to fire Gebru because her paper was insufficiently pro-Google, they damaged their whole purpose of having an ethical AI group. The demands of the group may seem childish or absurdly far-reaching, but if Google wants to rebuild their credibility here, this is the standard they have to hit. Otherwise, their own researchers have less credibility because their output will be viewed as "approved by Google management", not real research. This is critical if they need to say "our own AI ethicists say this is ok."