Consumer expert Martin Lewis has warned against using artificial intelligence for making important financial decisions such as pensions, investments or tax benefits without thorough verification.
AI for budgeting and pattern spotting
Speaking on his BBC podcast, Lewis said AI can be "absolutely brilliant" for budgeting, spotting patterns and identifying spending habits that users might not notice themselves. However, he cautioned against simply uploading bank statements into AI tools due to privacy risks.
He warned that data entered into AI systems is often used to train future models, meaning it is not private. Users should check whether the tool uses their information for training before sharing sensitive details.
Reading terms and conditions
Lewis described AI as "really, really useful" for reading through lengthy documents such as travel insurance policies. He advised users to verify the information by asking the AI to point to the specific section it used.
What to avoid with AI
"I definitely wouldn't use it to tell me what I should invest in. I definitely wouldn't use it for tax benefits or pension decisions without checking, and I definitely wouldn't, unless I knew it was a locked-in, secure environment, upload my entire financial life, because I think there is a risk of access to scammers," Lewis said.
He added that he was unsure how easy it would be for hackers to obtain financial details from an AI tool, or whether some tools are completely safe. He called the security risk a "legitimate consideration" for now.
Different levels of AI
Lewis noted that the quality of answers varies significantly between different versions of AI. "Most people are on the quick or the instant version, and if you're on the quick and the instant version, that is designed to give you a quick, instant, may I say somewhat flippant answer," he said.
He recommended using the highest level available, such as medium or pro versions, especially for important financial matters. "The entire research base and what it's checking and cross-checking itself is much, much more rigid," he explained.
Always double-check
Even with advanced AI, Lewis urged users to continue verifying results. "And you always, even then, want to still continue to double check," he said.
