My AI Agent Recommended a Non-Existent Investment Product — Exposing Information Gaps Between Fund Distributors and Asset Manage
Description and use
An automated fact‑checking recipe for an AI agent recommending investment funds, using primary data from asset managers and cross‑checking AI outputs to prevent fabricated or outdated recommendations.
Problem
An AI agent fabricated a non-existent investment trust recommendation, exposing gaps in information reliability and hierarchy between primary and secondary sources.
Outcome
The system now automatically flags AI recommendations that do not exist in the master data and detects discrepancies between distributor and asset manager information, preventing erroneous investment decisions.
Workflow
- 1. Define primary data source (asset manager official website).
- 2. Collect AI recommendation list.
- 3. Load primary data into a pandas DataFrame.
- 4. Load AI recommendation into another DataFrame.
- 5. Iterate over AI recommendations and check existence in primary data; flag missing entries.
- 6. Merge secondary source data (distributor sites) with primary data to detect discrepancies such as NISA eligibility.
- 7. Output warnings for any inconsistencies or fabricated products.
Evidence
The AI had confidently fabricated a plausible-sounding product and recommended it.
The NISA growth investment category eligibility status conflicted between the two sources.
- Record date: 2026-09-07T18:59:14.813Z