AI mistakes pile up in publishing, politics, and court filings

In short: More people are using AI for serious work, and public mistakes are showing up more often, including in books, media, and court documents.

What's going on

A Financial Times column argues that we should be getting better at using AI by now, but recent examples suggest the opposite. The piece describes a Canadian lithium company document that included a bizarre, nonsensical sentence, which the author suspects came from an AI writing tool.

The column also lists several public cases where AI use led to embarrassment or discipline. A US horror novel called Shy Girl had its release cancelled by its publisher after fears that AI helped write it. Journalists in several countries were reported to have lost jobs or been suspended over how they used AI in their work.

In another example, US media entrepreneur Steven Rosenbaum said AI created fake or wrongly credited quotes in his book The Future of Truth: How AI Reshapes Reality. The column suggests the temptation is easy to understand, since AI can save time, but it can also produce confident-sounding errors.

Courtrooms are another place where this is showing up. In one Oregon winery dispute, lawyers were ordered to pay $110,000 after citing made-up legal cases. A French lawyer and academic, Damien Charlotin, who tracks AI misuse in legal filings, said it was the largest penalty he had seen so far.

What to watch

Charlotin warns that “large language models hallucinate by default,” meaning they can invent details while sounding sure (like a student who writes an answer that looks neat but is not true). The key question is whether workplaces and courts will set clearer rules, and whether people will start treating AI as a draft helper, not a final authority.

Source: Financial Times