A recent corporate filing from a Canadian lithium explorer, preparing to list on the Australian Securities Exchange, suddenly veered into gibberish about shares being “ejaculated out” and a “helpful girlfriend” troubling local councils.
The passage was so bizarre it turned a dry governance statement into a viral curiosity instead of a regulatory footnote. It also raised a sharper question about how many more of these AI-powered blunders are slipping through unnoticed.
The lithium group, Li-FT Power, produced the corporate governance statement as part of its ASX debut documentation, a routine disclosure usually ignored by everyone except compliance teams and regulators. Instead, one mangled sentence grabbed attention worldwide, precisely because such filings are expected to be bland and tightly vetted.
An inquiry was sent to Li-FT Power asking whether artificial intelligence generated the offending line or how else the mess occurred. No response arrived before deadline, leaving a mystery over what systems, checks and tools the company used.
Observers find it difficult to identify any culprit other than a generative AI system, commonly referred to in shorthand as Chat, which is increasingly embedded in corporate workflows. These tools often autocomplete or rewrite text that staff then paste into official documents, assuming the content is safe.
When nobody reads the boilerplate closely, nonsense can slip through unchallenged and end up lodged in market filings and public records. The Li-FT episode joins a fast-growing catalogue of AI misfires surfacing in everything from marketing copy to legal submissions.
That catalogue is expanding so quickly it has already derailed plans to bundle the year’s most egregious AI mistakes into a neat end-of-year wrap. We are barely halfway through the calendar, yet the list of missteps already looks unwieldy and fresh ones appear almost weekly.
AI is rewriting more than company documents, it is rewriting expectations about how often technology will fail in places where precision supposedly matters most.
