When AI Fails

AI is making waves in compliance, but as tempting as it is to rely on AI for regulatory matters, the risks are undeniable.

Compliance isn’t just about finding the right answer. It’s about context, nuance, and judgement. AI can assist by drafting policies, summarising regulations, and flagging anomalies, but it doesn’t interpret grey areas, adapt to regulatory changes, or defend your firm in front of regulators - YOU DO!

Take a look at the screenshot I’ve shared below. When asked about RTS 22 transaction reporting, ChatGPT confidently gave an answer, and got it wrong. It incorrectly stated that the Execution Within Firm (Field 59) should be left blank when a firm “only transmits the order and does not execute it.”

However, regulatory guidance is clear: this field should be populated in every transaction report, reflecting the individual or algorithm making the execution decision. Since RTS 22 defines execution to include reception and transmission, the person or algorithm determining which firm to transmit the order to must be recorded here. If the determination is made by an external party (e.g. the client or an employee of a company within the same group), the field should be populated with NORE.

A minor mistake? Perhaps. But in compliance, precision matters, and errors like this could lead to misreporting, regulatory breaches, and hefty fines.

AI is a powerful tool, but it’s just that, a tool. Always double check its outputs against official regulatory guidance and internal policies. While technology is growing, we still need our human expertise to keep everything on track. Our jobs? Safe for now.

In the end, AI can make our work easier, but it can’t replace the critical thinking, judgement, and experience we bring to compliance.