Thanks to BBC Tech https://www.bbc.co.uk/news/articles/c743j83d8kzo for reporting the fact that a recent trial of AI by a major British law firm had very mixed results.
They put chatbots to the test by asking 50 relatively hard questions about English law, and the results ranged from hopeless to potentially handy.
Their conclusions were that chatbots were generally getting to the stage where hey could be useful for real legal work, but only with expert human supervision.
This reinforces what we always say, and which I hope is incorporated into your AI statement or policy, that the results must always be checked by a real person with suitable expertise.