FINRA officials are digging deeper into how artificial intelligence, like large language models, and social media are being used by broker-dealers to market to investors.
The industry self-regulating body is expected to review about 67,000 public communications of firms this year, said Amy Sochard, vice president of FINRA’s Advertising Regulation Department.
That “is what we’re forecasting for this year, which is considerably up from last year,” she said during FINRA’s advertising regulation conference in Washington, D.C., on Sept. 26.
The last time FINRA hosted its advertising regulation conference was in mid-2022, before OpenAI’s ChatGPT launched and well before most advisors began using such large language models to help draft emails, social media posts and other forms of client communications. A major concern is whether the information being put in these LLMs is a part of an OpenAI, public system.
READ MORE: In AI we trust? The peril and potential of embracing AI in wealth
“What we think might be a better way to go is to have that incorporated into sort of your in-house tool that that you’re using, and not going out to the public source,” Sochard said, referring to AI tools that help firms translate from a foreign language. “It’s a fascinating and fast-developing area. And I do think it’s one of the things that artificial intelligence will be valuable for.”
Where AI fits in FINRA’s marketing rules
In May, FINRA updated its Rule 2210 guidelines to include the use of chatbots and AI in communications with investors and the public. The largest takeaway from those guidelines, or FAQs, was that there needs to be clearly defined supervision procedures that comply with existing marketing rules.
“It simply says that if you use a chatbot or a large language model to create content, then it must adhere to the communications with the public rule standards, including the fair and balanced requirements and the prohibition on misleading statements,” Sochard said.
However, there are still many areas of AI that are not clearly defined in regulations. When asked how FINRA’s 2008 guidance on ghostwritten materials applies to AI-generated content, Sochard hesitated to give direction.
“It’s really interesting thinking about this with respect to AI. You’re asking me on the fly. I’ve never thought about this before,” she said. “I’d be kind of inclined to say, yeah, you do need to say, ‘I didn’t write this,’ especially if it has the appearance of something that you did write.”
FINRA leaders offer tips on using social media influencers
FINRA officials also discussed the results of recent so-called sweeps of firms that were paying social media financial influencers, called finfluencers, to promote an investment service or product and whether those firms were properly disclosing this type of advertising, which resulted in three enforcement actions this year.
READ MORE: FINRA, SEC send warning on deepfakes amidst their own AI plans
“We had 15 firms. We reviewed over 1,300 communications, and the overall, the noncompliance rate was about 70%,” said Stephanie Gregory, associate director of the complex review team in FINRA’s Advertising Regulation Department. “We found that over half, about 55% of them, did not disclose that the advertisements were paid. … And then we saw a lot of statements and claims, promissory, misleading. And for the most part, they all tended to be about investment success: how you can make a lot of money.”
FINRA officials also said some firms did not disclose the relationship between the firm and paid influencer as well as any other affiliated parties. And some firms were found to not have a written supervisory procedure for this type of social media marketing, and they weren’t retaining records, as required by regulators.
Firms “gave them [the influencers] talking points of things to say, but they didn’t supervise any of it,” Gregory said. “You’re responsible for all records of those communications. Make sure that you have or you’re capable of retaining them.”
Gregory also advised broker-dealers to make sure that any hired finfluencer is given education on the topic or product they will be discussing on social media.
“Firms should consider vetting influencers. Do your due diligence to try to avoid any compliance or regulatory risk down the line,” she said. “If they have a lot of followers, maybe additional controls should be in place.”
FINRA weighs guidance on using AI for language translation
FINRA leaders were also asked about how firms should handle client communications in foreign languages, especially when there are AI tools that can rapidly translate and dictate conversations or documents.
So far, FINRA officials have been talking with firms, foreign regulators and the U.S. Securities and Exchange Commission about how to create a broader roadmap of all the rules and guidance that exist in this area, Sochard said.
“We’ve been drafting a regulatory notice that would provide this sort of overall landscape picture of all the rules and guidance that already exist. Because there are things that have been out there, things from the SEC, there’s things from FINRA. So we want to get that out,” she said. “We also want to summarize some of these effective practices … and we also want to provide some resources that we’ve discovered along the way that we think might be of use to firms that choose to go down this path.”
Sochard, who plans to retire this November, said she hopes that FINRA can have this roadmap out before her retirement.
It’s “going through an internal FINRA review process, and most likely we’ll also want to circle back with some of the SEC staff on this as well,” she said. “Hopefully, this will be coming soon. I’d love to see it out before I retire but, you know, sometimes these things as we learn … can take a while.”