FINRA’s President and Chief Executive Officer, Robert Cook, recently published a blog post describing FINRA’s ongoing efforts to integrate artificial intelligence (“AI”), including generative AI (“GenAI”), into its regulatory program and engage with member firms regarding their use of GenAI. Below we highlight how FINRA is using GenAI to perform its self-regulatory functions more effectively and efficiently.
FINRA’s Adoption of AI Tools
FINRA has been developing and using AI to support its market oversight functions for years. For example, FINRA employs algorithms that use AI to review hundreds of billions of market events generated each day to identify potential fraud, manipulation, or other misconduct, which staff may investigate further or, for matters outside of FINRA’s jurisdiction, refer to the U.S. Securities and Exchange Commission.
According to Mr. Cook, FINRA approaches new AI-enabled tools carefully to ensure each meets its intended purpose and that potential risks are identified and addressed. In this regard, FINRA’s overall process is coordinated by an internal AI Governance Committee. FINRA maintains oversight of AI-enabled tools through security assessments and regular reviews to ensure the tools continue to meet FINRA’s security standards and safeguard regulatory information. In addition, FINRA has made foundational GenAI training available to all staff, together with more specialized training of different tools and how to use them.
Deployment of GenAI-Enabled Tools
Last year, FINRA introduced an internal large language model (“LLM”)-based chat tool, called “FILLIP.” Staff use FILLIP for support with various regulatory functions, including information summarization, document comparison, and drafting, as well as assisting with member firm risk reviews and analyzing data on mutual funds and exchange-traded funds to facilitate sales practice examinations. Other GenAI-enabled tools developed by FINRA include:
- A tool to summarize and analyze investor complaints filed directly with FINRA (which, according to Mr. Cook, in recent years have exceeded more than 10,000 per year), which enables faster review and response by investigators;
- A tool to summarize and analyze comments on FINRA Regulatory Notices to aid in more quickly understanding and considering feedback on rule proposals;
- A tool to summarize and evaluate disclosures filed with FINRA’s Credentialing, Registration, Education and Disclosure (CRED) department; and
- A tool to extract and compare information from firms’ eFOCUS reports.
Collectively, FINRA expects these initiatives to yield considerable annual efficiency gains, allowing FINRA to better serve the interests of investors and member firms.