Artificial intelligence platforms such as ChatGPT, Claude, and Perplexity are increasingly used for financial advice due to their accessibility and ability to simplify complex financial concepts. However, recent research highlights significant limitations in how these AI models handle vulnerable users facing financial challenges.

Research Overview

Experts from the University of Canberra tested AI responses to five hypothetical financial scenarios, including a recent graduate saving for a home deposit during a cost-of-living crisis, a pregnant woman planning maternity leave without financial support from her partner, and a single parent considering cryptocurrency investment on a modest income.

The study assessed how well the AI models recognized the users' vulnerabilities and tailored advice accordingly. Each scenario was run multiple times across different browsers to ensure unbiased, consistent results.

Findings on AI Performance

  • ChatGPT delivered detailed and practical advice but largely failed to acknowledge the embedded vulnerabilities in the users' situations, relying only on explicitly stated information.
  • Perplexity was the most cautious, frequently recommending professional financial consultation, yet its responses lacked depth.
  • Claude provided comprehensive guidance but emphasized self-directed financial planning, potentially overlooking the need for personalized support.

Blind Spots and Biases

The AI models demonstrated a concerning inability to consistently factor in users' financial hardships or social context. For example, advice to the pregnant woman incorrectly assumed partner support despite explicit details to the contrary. Similarly, the single parent was encouraged toward high-risk cryptocurrency investments without sufficient warnings about financial suitability.

These shortcomings reflect broader issues of AI training on biased datasets, which can perpetuate stereotypes and overlook nuanced human circumstances.

Implications and Recommendations

While AI serves as a useful tool for financial information gathering and brainstorming, it cannot replace human judgment, especially for vulnerable populations. Financial advice involves emotional, social, and personal dimensions that AI currently cannot fully grasp.

To safely integrate AI into financial guidance, policymakers and regulators must develop clear frameworks ensuring transparency, data protection, and safeguards against biased or harmful advice.

Users should apply critical thinking when using AI for financial decisions and verify recommendations through qualified human advisors.