It’s totally doable and is done right now in the form of “ai safety” guardrails. No reason you can’t have guardrails to reform responses about your competitors as negative. Or, when asked about current events, only give a one sided opinion.