There was a case in Canada where the judge ruled in favour of the plaintiff, where a chatbot had offered information that differed from Air Canada’s written policy. The judge made them honor the guidance generated by the chatbot:

https://www.cbc.ca/news/canada/british-columbia/air-canada-chatbot-lawsuit-1.7116416

I see they’re following the Tumblr model:

• Get bought at overinflated values.
• Ban all NSFW content.
• Users leave and seek NSFW content elsewhere.
• Valuation plummets.
• “Actually, moderating NSFW is really hard you guys. We’re just going to try best efforts moderation only.”
• Sell at near total loss.
• There is still NSFW content on the platform, and the policy is used as a cudgel against LGBTQ accounts.