The other day I was alerted to a company document that should have been so fantastically dull it was read by almost no one.
Alas, the corporate governance statement of a Canadian lithium explorer ahead of its Australian stock exchange debut was widely read last month, because one of its sentences said this: “You will change in share are all ejaculated out your helpful girlfriend issue she asked the councils responsible.”
I emailed the lithium company, Li-FT Power, to ask if AI was to blame and, if not, how the botching happened. But by deadline, there came no answer.
It is hard to think of any culprit except what many now refer to as Chat, which has made me rethink a plan I hatched in January to write a column in December looking back at a year of AI blunders.
READ MORE
It seems safer to do a half-year update since it is only June and the hall of AI shame is already overflowing with exhibits.
They remind me of something a tech CEO told me about large language models at the start of this year: businesses are racing to use a technology that was not chiefly designed for businesses to use.
Companies are far from alone.
Few weeks have passed this year without a politician, journalist or author being outed, sacked, questioned or ridiculed over AI. Resistance has been more or less futile.
Former Reform UK candidate, Matt Goodwin, insisted that multiple claims that AI had written parts of his book on immigration, Suicide of a Nation, were categorically untrue. This did not stop him acquiring a nickname I feel sure he will struggle to lose: MattGPT.
Labour MP, Mike Reader, is still known as the ChatG-MP months after reports that he had been spotted on a train using Chat to respond to constituents.
We must live in a world where AI is being overused and overtrusted by too many who should know far better
It could have been worse. In March, the US release of a horror novel, Shy Girl, was cancelled by its publisher because of fears AI was used to help write it.
Meanwhile, news of journalists losing their jobs or being suspended over AI use emerged from New York to the Netherlands in February, March and April.
May brought an even more startling development.
The American media entrepreneur, Steven Rosenbaum, admitted AI had dreamt up a number of fake or misattributed quotes in a praise-larded book with the hard-to-make-it-up title of The Future of Truth: How AI Reshapes Reality.
I find it easy to imagine how a time-pressed writer or politician might find the allure of AI irresistible. But the growing degree of AI slop in courtroom legal citations continues to dismay, as does the size of the fines that frustrated judges are now imposing as a result.
In March, it emerged that lawyers who had cited multiple bogus cases in a family dispute over an Oregon winery had been ordered to pay $110,000 (€95,458) in penalties.
“In the quickly expanding universe of cases involving sanctions for the misuse of artificial intelligence, this case is a notorious outlier in both degree and volume,” wrote magistrate judge Mark Clarke.
Damien Charlotin, a 34-year-old French lawyer and academic who keeps a database of AI legal misuse dating back to 2023, told me the winery case penalties were the highest he had seen so far.
I doubt they will be the last. Since AI took off, the number of litigants going to US federal civil courts without a lawyer has jumped from 11 per cent to 17 per cent. I feel confident they will swell court backlogs, as well as Charlotin’s database. It listed fewer than 700 cases in December but as of this week it had more than 1,500.
That is not so high in the overall scheme of things, says Charlotin, who teaches automation, writes computer code himself and likes the extra time AI gives him to focus on the human parts of his work.
But he also understands that, as he puts it, “large language models hallucinate by default”. They need to be handled with care and, in many cases, reduced to specific tasks. In other words, humans don’t just need to be in the loop. They need to stay firmly in charge.
Eventually, this may become more obvious than it seems to be today. In the meantime, we must live in a world where AI is being overused and overtrusted by too many who should know far better. – Copyright The Financial Times Limited 2026
