FRM Part II · FRM Exam Part II · Advances in Artificial Intelligence: Implications for Capital Markets Activities
A risk manager notes that a generative AI tool used to summarise regulatory changes sometimes produces confident but fabricated citations. Which risk control is most appropriate before relying on its output for compliance decisions?
The firm should require human verification of the tool's citations against original regulatory documents and keep logs of outputs. Generative models can produce fluent but fabricated references, so fluency or self-reported confidence is not reliable evidence of accuracy. Making the tool more creative would increase the risk.
- ARequire human verification of cited sources against original documents, with logging of outputsCorrect
- BIncrease the tool's creativity setting to broaden coverage
- CAccept outputs that sound authoritative without review
- DLimit review to outputs the tool itself rates as low-confidence
Explanation
Fabricated outputs, or hallucinations, are a known generative AI risk. Human verification against original sources, with logging for audit, directly mitigates it. Fluent wording is no evidence of accuracy, and the model's self-rated confidence is unreliable.
Did you get it right without looking?
One question tells you little. A timed set on Advances in Artificial Intelligence: Implications for Capital Markets Activities shows your real accuracy, how long you take and where you lose marks.
More Advances in Artificial Intelligence: Implications for Capital Markets Activities questions
- An asset manager uses a machine learning model to allocate client orders and discovers that, because of biased historical data, it systemati…
- A bank trains a machine learning credit-spread forecasting model on ten years of data covering a calm, low-volatility market regime. Shortly…
- A risk team uses a complex machine learning model to flag potential market manipulation. A regulator asks the firm to explain why a specific…
- Several large asset managers license the same foundation model from one provider to generate trading signals. A regulator warns that this ra…
- A bank trains a machine-learning credit-trading model on ten years of market data in which volatility was persistently low. During a sudden …
- A quantitative trading firm uses a deep reinforcement learning agent to execute large equity orders. Backtests show strong results, but in l…