GILES RAFFERTY, Corporate Communications and Media Advisor
‘Do Androids Dream of Electric Sheep?’ is the title of a dystopian science fiction novel by Philip K Dick published in 1967. Fast forward to 2025 and we discover that Ai Chatbots are prone to bouts of ‘hallucination’. IBM says Ai hallucinations are when a chatbot perceives patterns or objects that are non-existent or imperceptible to human observers, creating outputs that are nonsensical or altogether inaccurate.
The British Broadcasting Corporation, the BBC, published research, in February of this year, into Ai Hallucinations. It found that AI Chatbots introduced “significant inaccuracies” when summarizing news content. Something we at FIRST Advisers encountered during the recent July and August reporting season.
Ai hallucination to real world misinformation
It was quite a surprise when our media monitoring revealed a Tipranks story about how a client was facing ‘significant risk’ and ‘compliance challenges,’ when there was no increased risk or compliance burden. Days later it was just as surprising to be informed by AInvest that a completely unrelated 3rd party, was offering a completely unrelated cash flow management solution to its customers that was somehow driving the same client’s share price higher. It was jolly nice of them to do so but, alas, this story was also complete nonsense.
Both Tipranks and AInvest are Ai enabled retail investor platforms. Both offer limited access to free content as a lure to get retail investors to pay a subscription for greater levels of access. And both platforms carried Ai generated stories that were at best misinterpreting our clients ASX filings and at worst conflating completely unrelated stories to generate a fundamentally misleading article.
If you are prepared to work for it, you can find a disclaimer at the bottom of the AInvest webpages that admits – “The news articles available on this platform are generated in whole or in part by artificial intelligence and may not have been reviewed or fact checked by human editors.”. While Tipranks’ Ai content has a Tipranks Auto-Generated News Desk byline and again, if you are prepared to work for it, you can click on the byline and get to an About page which it claims Tipranks uses AI to scan company announcements, translating complex ideas into easily understandable articles.
Keep a human hand on the tiller
There is a sense of inevitability about the adoption of Ai to accelerate and streamline content generation. But as long as chatbots are prone to ‘dream of electric sheep’ there is also a sense of inevitability around inaccurate, misleading and potential harmful Ai fabrications being peddled as news. Research by NewsGuard, an organization tracking misinformation, found at least 49 ‘news’ websites that are “almost entirely written by artificial intelligence software”.
In our increasingly automated world, there may be little that can be done to correct erroneous Ai stories. Thankfully they are occurring, for now and for the most part, at the very periphery of online ‘news’. The best course of action will likely be to make sure to have effective media monitoring in place, that is curated by humans, to avoid surprises and set the record straight in the event an Ai hallucination prompts misinformed questions.