I have found all of the AI hazards cited in the video to be true. I've also found AI tools over-eager to please the user with flattery. I've had to prompt AI not to be a "yes-man". Otherwise, it just agrees with me and lauds my brilliance 🙄. I also recently discovered that Claude, at least, uses a command line tool instead of a pdf reader to read pdf files and that it garbles the pdf's content. It cited as errors these garbled passages when, in fact, they were accurate. AI tools warn users not to take their responses at face value. As authors, we have to fact-check AI or hire human fact-checkers. I've written a lengthy chapter on AI in my upcoming book, Find Your Cloud 9s With SaaS Apps. Besides my own personal experiences with AI accuracy, I noted that the global research agency, Deloitte, relied too heavily on AI in a report for an Australian government agency. A researcher discovered fabricated quotes, footnotes, and references to a non-existent book and external reports. Deloitte had to correct the errors, reprint the report, and refund part of its fee.