SEBI’s circular on reporting for AI and ML applications and systems offered and used by mutual funds
The Securities and Exchange Board of India (SEBI) has asked all mutual funds companies in India to report Artificial Intelligence (AI) and Machine Learning (ML) applications and systems offered to investors (individuals and institutions) or used internally by it to facilitate investing and trading or for any other purpose. The move is aimed at conducting a survey for creating an inventory of the AI / ML landscape in the Indian financial markets to gain an in-depth understanding of the adoption of such technologies in the markets and to ensure preparedness for any AI / ML policies that may arise in the future.
“As most AI / ML systems are black boxes and their behaviour cannot be easily quantified, it is imperative to ensure that any advertised financial benefit owing to these technologies in investor facing financial products offered by intermediaries should not constitute to misrepresentation,” said SEBI in its circular dated 9 May addressed to all mutual funds, asset management companies, trustee companies, board of trustees of mutual funds and Association of Mutual Funds in India.
Key highlights of the circular
- All registered mutual funds offering or using applications or systems should participate in the reporting process by completing the AI / ML reporting form.
- With effect from quarter ending June 2019, these entities are required to fill in the form and make submissions on quarterly basis within 15 calendar days of the expiry of the quarter to AMFI.
- Other than the basic details of the entity, the form also seeks details such as-
- What claims have been made regarding AI and ML application / system?
- How was the AI or ML project implemented?
- Does the system disseminate investment or trading advice or strategies?
- Is the application / system used in area of cyber security to detect attacks?
- What safeguards are in place to prevent abnormal behaviour of the AI or ML application / System?
- AMFI shall consolidate the information on AI / ML applications and systems reported by mutual funds on quarterly basis and submit to SEBI within 30 calendar days of the expiry of the quarter, starting from quarter ending June 2019.
- SEBI has issued an indicative list of systems deemed to be based on AI and ML technology. It includes the following-
- Natural Language Processing , sentiment analysis or text mining systems that gather intelligence from unstructured data. In this case, voice to text, text to intelligence systems in any natural language will be considered in scope. E.g., robo chat bots, big data intelligence gathering systems.
- Neural Networks or a modified form of it. In this case, any systems that uses a number of nodes (physical or software simulated nodes) mimicking natural neural networks of any scale, so as to carry out learning from previous firing of the nodes will be considered in scope. E.g., recurrent neural networks and deep learning neural networks.
- Machine learning through supervised, unsupervised learning or a combination of both. In this case, any application or systems that carry out knowledge representation to form a knowledge base of domain, by learning and creating its outputs with real world input data and deciding future outputs based upon the knowledge base. E.g., system based on Decision tree, random forest, K mean, Markov decision process, Gradient boosting algorithms.
- A system that uses statistical heuristics method instead of procedural algorithms or the system / application applies clustering or categorization algorithms to categorize data without a predefined set of categories.
- A system that uses a feedback mechanism to improve its parameters and bases it subsequent execution steps on these parameters.
- A system that does knowledge representation and maintains a knowledge base.