FINRA Guides Firms to Build Investor Protections Around AI Tools

Burgeoning use of generative artificial intelligence has touched nearly every industry, and financial services are no exception. Questions are brewing around the contours of regulating these innovations.

While AI use increases, the Financial Industry Regulatory Authority recently updated its advertising regulation FAQs offering compliance guidance for how broker-dealers can use AI chatbots and how to approach AI-created communications. FINRA often uses FAQs to guide use of emerging technologies absent specific regulation, or where formal rules for narrow use cases may not materialize. One crucial throughline of these FAQs is FINRA’s focus on investors—the intended audience for AI chatbots and AI-generated content.

For compliance teams, this focus essentially serves as a brightline rule. Whenever AI is deployed in a way that touches or interacts with investors, whether in the form of a chatbot or AI-generated content that’s shared with an investor—the constellation of FINRA rules applies.

Guidance

First, FINRA tackles supervision of AI chatbot communications with investors. From an oversight perspective, the applicable rules for AI chatbots track FINRA’s existing standards for retail communications, institutional communications, and correspondence under Rule 3110—firms must ensure content shared through AI chatbots meets relevant pre- or post-review regulatory mandates. FINRA also stresses that AI chatbot content itself mustn’t contain “unwarranted, promissory, or misleading” statements, among the other requirements under Rule 2210(d).

Regarding the compliance requirements for AI-generated content, FINRA again takes the position that such content is subject to the same supervisory requirements as any other content shared with investors. As with chatbots or any human-created content, the review protocols outlined in Rule 3110 apply as do the standards for good faith and fair dealing under Rule 2210(d).

The guidance for AI chatbots and AI-generated content doesn’t depart from existing FINRA compliance standards. Firms hoping that use of AI robot friends for generating certain types of content might exempt them from certain FINRA compliance requirements will be disappointed in these FAQs. That said, FINRA’s approach here is squarely in line with its guidance on issues such as social media and collaboration platforms, which grafted existing compliance expectations onto new technologies to maintain a consistent compliance stance and serve the ultimate purpose of protecting investors.

As a threshold matter, firms must establish policies and procedures for deployment and use of AI chatbots and develop guardrails for how employees can use generative AI tools to create investor content.

These policies might address procurement protocols for selection of new AI technologies, considerations for implementation, and more granular expectations about AI governance—training, maintenance, and oversight of AI to comply with FINRA’s content, supervision, and recordkeeping requirements.

Compliance teams must:

• Consider technical and supervisory controls for oversight of these new AI capabilities
• Coordinate with data scientists to validate that AI training and testing are sufficient to minimize risk of potentially problematic interactions, such as providing incorrect information, or hallucinations, during AI chatbot interactions
• Contemplate deploying supporting technologies to flag the presence of a chatbot during or after a meeting to ensure its participation is transparent to investors and alert surveillance teams conducting subsequent reviews that an AI assistant may have generated content during the meeting

Chatbots and AI-enabled assistants are now part and parcel of collaboration applications such as Microsoft Teams, Slack, WebEx, and Zoom. AI assistants generate feedback during meetings and create post-meeting summaries.

Compliance officers should ensure AI chatbots and other AI tools that generate content provided to investors are configured with recordkeeping capabilities. The goal is to capture and retain relevant portions of conversations or interactions to comply with FINRA, the Securities and Exchange Commission, and the Commodity Futures Trading Commission Rules.

Regarding AI-generated content, firms may need to augment or improve technologies for pre-review of retail communications, or post-review of institutional communications and general correspondence. This can ensure better compliance with FINRA Rule 2210 and other rules for disseminating information to investors.

Although firms may use existing toolsets to manually or automatically review content, new approaches to content vetting using AI itself may be beneficial. For example, while words such as “guarantee” or “promise” may present issues for a business in certain contexts, they may be acceptable in other use cases, such as “guaranteed return” versus “guaranteed friendly” services.

AI-based review technology can assist with:

• Analysis of investor content by analyzing the context of a communication to determine if it’s appropriate, or not.
• Reviewing and identifying potential risks in investor content, which may span multiple languages. Critical to these AI-enabled content vetting processes is the injection of a “human-in-the-loop” to validate that risks or issues flagged using machine learning are accurate and valid.

Use of AI-powered chatbots and AI-created investor content is likely to become the norm. Embracing these tools and making related investments in compliance processes, technologies, and teams in a manner consistent with FINRA’s expectations will set your firm on a path to success.

This article does not necessarily reflect the opinion of Bloomberg Industry Group, Inc., the publisher of Bloomberg Law and Bloomberg Tax, or its owners.

Author Information

Marc Gilman is general counsel and vice president of compliance at Theta Lake, and adjunct professor at Fordham University School of Law.

Write for Us: Author Guidelines