In using AI / ML growing in exchanges, brokers, and mutual funds, Sebi intended to strike a balance between change and regulation. The proposed Table Table contains management, data security, biased prevention, investor protection, and risk control.
The last date to send comments / suggestions is July 11.
Previously, Sebi consists of a working group to study Indian, Global Best AI / ML attributes, which it rules in preparing the instructions for use in AI / ML applications. The working group is also driven that provides recommendations to resolve concerns and issues related to using AI / ML applications.
Read again: SEBI can lighten management rules for PSUs with more than 90% government hold
Here are 5 principles guiding:
1) Model management students must set internal teams with technical skills to monitor AI / ML models, and handle exceptions. Management Management is accountable for the whole life of AI, including managing third party sellers.
2) Mandatory Revelation
If the AI / ML appliances directly impact investors – such as advery or advisory-firms advisor should express their use, including models, risks, and limits, and limits. Language should be simple and friendly to client.
3) Strong attempt test and monitor
Sebi suggests more modeling tests in simulated environments before live deployment, and continuously monitor after. Companies should keep data logs data and documentation for a minimum of five years to ensure justice and traceability.
4) Reason and bias
To prevent discrimination, Sebi suggested to use different quality data and training staff to determine bias.
5) Data Security
Companies have to follow strong data management, privacy, and protocols in cybersecure.
Read again: SEBI Board Meeting: Regulator approved PSU delisting, IPO reform, deferring securities. 10 Key Taaway
Sebi suggests a lighter regulatory method for models available in the room (eg, monitor), while models affecting clients will face clients to face clients.
(Disclaim: Recommendations, suggestions, views and opinions given to experts themselves. It does not represent economic views
