Securities and Exchange Board of India (SEBI) on Friday proposed guiding principles for responsible usage of Artificial Intelligence (AI) and Machine Learning (ML) applications in securities markets to safeguard investors and market integrity.
Also, the regulator has proposed that a "regulatory lite" framework may be adopted for usage of AI/ML in the securities market for any purpose other than for business operations that may directly impact their customers.
The proposed "guiding principles are intended to optimise benefits and minimise potential risks associated with integration of AI/ML-based applications in securities markets to safeguard investor protection, market integrity, and financial stability," SEBI said in its consultation paper.
At present, AI/ML is being used by market participants mainly for advisory and support services, risk management, client identification and monitoring, surveillance, pattern recognition, internal compliance purposes and cyber security.
"While AI/ML has the potential to improve productivity, efficiency and outcome, it is also important to manage these systems responsibly as their usage creates or amplifies certain risks which could have an impact on the efficiency of financial markets and may result in adverse impact to investors," SEBI said.
Accordingly, SEBI proposed high-level principles to provide guidance to the market participants for having reasonable procedures and control systems in place for supervision and governance of usage of AI/ML applications or tools.
The proposed guiding principles were suggested by a SEBI-constituted working group after studying the existing AI/ML guidelines in India as well as globally.
SEBI working group’s suggestions
As a part of the proposal, the working group suggested that market participants using AI/ML models should have an internal team with adequate skills and experience to monitor the performance, efficacy and security of the algorithms deployed throughout their lifecycle, as well as maintain auditability and explain interpretability of such models.
Furthermore, the team should establish procedures for exception and error handling related to AI/ML-based systems.
It should also establish fallback plans in the event an AI-based application fails due to technical issues or an unexpected disruption to ensure that the relevant function is carried out through an alternative process.
It has been proposed that market participants using AI/ML models for business operations -- such as selection of trading algorithms, asset management or portfolio management and advisory and support services -- that may directly impact their customers should disclose the same to the respective customers to foster transparency and accountability.
The market participants should adequately test and monitor the AI/ML-based models to validate their results on a continuous basis. Further, it has been proposed that the testing should be conducted in an environment that is segregated from the live environment prior to deployment to ensure that AI/ML models behave as expected in stressed and unstressed market conditions.
Also, market participants should maintain proper documentation of all the models and store input and output data for at least 5 years.
"Since the AI/ML systems are dependent on collection and processing of data, market participants should have a clear policy for data security, cyber security and data privacy for the usage of AI/ML based models," SEBI said, adding that information about technical glitches, data breaches shall be communicated to it and other relevant authorities.
SEBI has sought public comments till July 11 on the proposals.