Compliance Risk

There are two steps:

(1) raw social data collection; and (2) analysis with NLP or GenAI

(1) is the difficult part. You would need to know where the discussions are happening for your client or its industry – is it in the news, Facebook, Instagram, a forum? Once you identified the data source, you would need to find a way to collect the data, and the legal and cost question comes in to play.

I recommend you to buy data from a third party (there are many of them). They can either provide you only the raw data feed (e.g. give me all the Facebook discussions about my brand), or they can provide you with a dashboard with standard analytics. If you only need one off data analysis for maybe a proposal, getting raw data and crunch it yourself is a cheaper, better way forward

For Canada specifically, please refer to the latest Draft Guideline E-23 on model risk management by OSFI (our banking and insurance regulator).

To capture the risk posed by AI, the Draft Guideline modernizes the definition of “model” in the original 2017 Guideline to explicitly include AI and machine learning methods. Basically the AI-based solutions are considered “models” and are subject to model risk management requirements. Detailed requirements can be found in the guideline