Regulatory Considerations for Artificial Intelligence Programs Providing Advice to Individually Managed Accounts
July 7, 2017
Artificial intelligence (AI) is expected to revolutionize many areas of modern life, including the investment management industry. AI applications currently are being used to develop management programs for investment funds, for example. One area where AI is not yet widely used, but where WSGR expects to see new AI applications in the future, is to support and implement more complex, individually tailored, and responsive asset allocation services based on a wide range of new and more quickly updated data sources.
For example, AI could be used to provide ongoing recommendations to individuals based on a combination of information provided to the adviser by the investor and data collected from an investor's online activities. In this context, AI could, among other things:
- Use historical and current pricing of securities and other assets to create refined trading algorithms
- Monitor and evaluate the investment impact of an investor's buying, social media, and other internet habits (for example, in terms of investment interests or a user's financial ability to make investments)
- Monitor and evaluate the investment impact of general social media trends (e.g., events trending on Twitter)
- Monitor bank accounts, investment accounts, and retirement accounts to recommend particular investment timing and amounts
- Identify major life events that may change investment considerations (e.g., a new job, marriage, house or child, or an investor nearing retirement age)
- Identify and analyze a broader range of potential investments in mutual funds, exchange-traded funds, publicly traded securities, private securities, and other instruments than is typically covered by a single advisory firm
- Generally incorporate a broader range of investor preferences than is typically analyzed in investment programs
- Respond to changes in the market and investor preferences in ways that are faster and more comprehensive than is possible in existing advisory programs
The development of AI-based individually managed accounts will raise significant regulatory issues, including, among others, regulation under the Investment Advisers Act of 1940 (Advisers Act) and similar state laws. Below, we address several significant issues under the Advisers Act for advisers that use AI in two scenarios: (1) to make recommendations that investors then must approve or reject; or (2) to invest client assets at the adviser's discretion based on the investor's predefined preferences. WSGR will address issues related to broker-dealers who run these programs in a future alert. Advisers providing these programs should also consider issues related to intellectual property, technology licensing, and others.
Most advisers that sponsor AI-based separately managed accounts will need to register as investment advisers under the Advisers Act or similar state law, and as a result will need to comply with regulations in the areas listed below, plus others:
- Advisory Contracts. Advisers using internet-based AI-based programs will need to develop online mechanisms for investors to view and execute their advisory agreements and to view the adviser's disclosures regarding material terms and risks of the advisory program.
- Advertising. The anti-fraud provisions of the securities laws prohibit, among other things, the use of testimonials in advertising, false or misleading claims (including "puffery"), and using performance information without significant related disclosures. These issues can be complex in the context of social media. For instance, the SEC views testimonials as including "likes" and comments by advisory clients or other users on Facebook or similar platforms. Similarly, because of the disclaimers required in performance-based advertising, an adviser may be restricted in its use of Twitter or similar platforms with character limitations.
- Custody of Client Assets. An adviser that touches an investor's money needs to comply with requirements related to where and how it holds a client's cash, securities, and other assets; mandatory audits of client accounts; when and how the adviser provides reports to investors; and many other issues.
- Execution of Client Transactions. Advisers can either pick a broker that executes securities transactions for their clients or work with investors' existing brokerage firms. Advisers have a responsibility of best execution in working with any brokerage they choose. When contracting with outside brokerage firms, advisers using AI will need to adequately protect their rights to data collected and generated by their AI programs.
- Compensation. Advisers may not earn compensation based on the capital appreciation of investor assets unless an investor is a "qualified client," which for natural persons means a person with at least $2,000,000 in net worth or $1,000,000 in investments with the adviser. Advisers must also be careful not to take any compensation based on the size or successful completion of any securities transactions, which may require broker-dealer registration. On the other hand, fees based on assets under management may not be meaningful depending on how recommendations are provided to investors and may not be well received.
- Confidentiality. Advisers are generally required to protect individual investors' personal non-public information, and must take reasonable care to protect against hacking and similar breaches. In an AI-based program, advisers may have access to, store, and create information about investors that is much deeper and more extensive than a fund adviser or a typical adviser to individuals may have, and as a result will need to take particular care to protect customer information. Given how critical the information may be to a consumer, a single hacking incident could cause significant damage to the adviser's reputation, and may completely undermine the adviser's business.
- Privacy. Advisers must adopt privacy policies and provide notice to clients of their information-sharing practices. Advisers that sell sensitive investor information will need to comply with notice and consent provisions regarding information sharing and should consider whether investors will object to the sale of information created about them by the adviser's AI programs.
- Fiduciary Duties. Advisers have a duty to act in the best interest of their clients and must disclose and/or mitigate conflicts of interest related to their management of client accounts, including conflicts related to allocation among investor accounts, the receipt of fees (including from broker-dealers or based on sales of customer data), and others.
- Supervision. Advisers are responsible for regulatory compliance by their employees and anyone who provides advice on their behalf, and must adopt policies facilitating this supervision, including a code of ethics designed to mitigate insider trading and other fraud. An adviser's code of ethics must also require employees to report on their securities trading activities. If an adviser's individually managed account program involves human interaction regarding the results of the program—for example, by providing investors the ability to discuss their investments with a representative of the adviser—supervision will be particularly important. In addition, advisers will have significant responsibilities related to their supervision of coders.
- Licensing Relationships. A developer of AI-based management programs that licenses or white-labels its technology to other sponsors—such as an investment bank or a registered investment adviser that provides its customers access to the program—may not have a registration requirement under the Advisers Act, depending on how the developer is paid, its visibility on the platform, and other considerations. Any licensing agreement, though, would need to be carefully crafted to appropriately protect the adviser's property rights over the AI technology, data collected by the technology, and data generated by the AI-based program.
For more information about AI-based investment management programs or any related matter, please contact Robert H. Rosenblum (202-973-8808, firstname.lastname@example.org); Susan Gault-Brown (202-973-8809, email@example.com); or any member of the fintech regulatory practice at Wilson Sonsini Goodrich & Rosati.
This WSGR Alert was prepared by Amy Caiazza.