Artificial intelligence (AI) can create ‘dangerous’ errors when answering pension questions, potentially leading people towards scams, Quietroom development lead, Joe Craig, has warned.

Speaking at the PLSA Annual Conference, Craig highlighted how pensions communications is no longer what the industry wants to ‘broadcast’, due to the new communications system powered by AI.

“If you ask Google what to do with your pension, AI will answer that. This can create a different type of error to the ‘wild hallucination’ inaccurate answers generated by AI that is often discussed,” Craig stated.

“It gives a more dangerous type of error, which may well be more technically accurate but one that we would immediately recoil against as it is wildly inadvisable… it pretty quickly says to take it all out as cash. It then steers you to the next question, ‘do you want to know how to do that’, so it takes just one click to guide you in that direction,” he explained.

This risks steering people towards unscrupulous advisers, putting people on the path towards a scam, Craig warned.

According to a poll of conference attendees, 14 per cent consider AI to big the biggest risk for the pensions sector, with 72 per cent stating that the use of AI is about ‘balance’ and 13 per cent considering it an opportunity. Sixty-nine per cent believe that AI will have a meaningful impact on the pensions sector, with 27 per cent predicting it to be ‘revolutionary’.

---