The Book on How NOT To Use AI Is Being Written Right Now

European payment network Klarna was one of the first fintech companies to go all in on AI. They were even featured as a case study by OpenAI. Then after a flood of customer complaints, they tried to quietly backtrack and hire back many human assistants. Recently an author was called out after readers discovered that the text of AI prompts was published in her final book.

A lawyer was caught using fake citations hallucinated by AI and failed to notice them. A tech company's AI coding tool deleted their full code database ... and then lied about it. This week, Hertz came under fire for using AI systems to assess rental car returns for damage and erroneously charging hundreds of customers for minimal or non-existent damages.

Altogether these stories, and many others to come are offering the fodder for an emerging collection of the ways to NOT implement AI. The frequency of these negative examples may be even more useful than positive tips ... but the evolution reminds me of the famous line from the second Jurassic Park movie where park creator John Hammond justifies sending another team back to the dinosaur island by telling Ian Malcolm that they are not making the same mistakes twice.

"Yeah," he says, "you're making all new ones." That's the story that is likely to keep unfolding with these AI implementations as well.