10 minute read | December.12.2024
SEC officials have declared artificial intelligence (“AI”) “the most transformative technology of our times” while cautioning that “if a public company is using AI, that company has to be honest about the role AI plays in its business and not exaggerate it to the point of AI washing[1].” To address this concern, the SEC has released guidance on AI washing, conflicts of interest, systemic risk, and fraud related to AI. This guidance has coincided with increased AI-related enforcement actions as part of the SEC’s “war” on AI fraud, and also appears to be influencing SEC comment letters.
A sample of the SEC’s disclosure comments issued since 2021 found at least 92 separate comments addressing AI-related disclosures, spanning comment letters issued to 56 different companies. This trend emphasizes the importance of considering the SEC’s AI-related guidance, as it continues to inform the SEC’s disclosure review process. By addressing this guidance in advance, companies may avoid undue scrutiny and improve the transparency and reliability of their AI disclosures.
Below is a summary of these disclosure comments, categorized by the key issues the SEC has raised through its guidance and enforcement actions as of October 2024.
As a threshold matter, companies must determine whether AI is material enough to require disclosure in their SEC filings. The SEC has advised companies to assess if discussions about AI in board meetings, earnings calls, and investor presentations suggest materiality and, if so, to provide corollary disclosures in SEC filings. Around 10%[2] of the SEC comments we reviewed addressed this issue of materiality, with the SEC indicating that as part of their materiality assessment, companies should consider evaluating, among other things, the impact of AI on their financial performance and growth, potential barriers to adoption, possible adverse developments and their implications, the evolving AI regulatory landscape, and the relevance of the referenced AI to the company’s business. Examples of such SEC comments[3] include:
We note disclosure regarding the importance of AI to your business, including that your financial performance and growth will be driven in large part by the demand for AI workloads. Please revise your business section to more fully discuss the current state of AI and the potential obstacles to broad-based AI adoption. In addition, more fully discuss the current state of AI regulation within the United States and your other markets. Clarify what you would consider to be “unfavorable developments” with the potential to materially impact the company, as referenced on page XX.
We note your disclosure that increased AI-driven advancement includes “a strong potential for interactive virtual assistants . . .” Please clarify whether you have plans to develop interactive virtual assistants, or otherwise, and please explain how this is relevant to your business.
Interestingly, the SEC has also shown concern about immateriality, requesting companies to justify the inclusion of certain AI-related disclosures that do not seem material and to include examples of use cases that would be helpful for investors’ understanding:
Please tell us why you believe that the various AI-related programs in the timeline table disclosed are material enough to be included, especially considering the early stage of such programs and the lack of disclosure regarding these programs.
While we note your disclosure about the ability of AI to predict the answers to frequently asked questions, it is not fully evident how AI and Machine Learning are used elsewhere in your business. Please revise to describe in greater detail, providing examples where useful to investors.
The SEC emphasizes that claims about AI prospects should have a reasonable basis, and this basis should be disclosed to investors. The SEC urges companies to avoid overstating their AI capabilities and making misleading disclosures about AI use or functionality if they cannot be supported. The SEC also encourages the inclusion of disclosure about the contemplated timing of any referenced AI that are in development or that has not yet been implemented. Approximately 30% of the SEC’s comments we reviewed addressed unsupported or unqualified statements. Examples of such SEC comments include:
We note the following disclosure on page XX: “We believe the Company is well positioned to capitalize on this trend by offering a unique and cutting-edge solution that leverages AI. By tapping into this growing market and delivering a superior product, the Company aims to become a leader in the conversational AI industry.”
Please provide additional details regarding how AI allows you to “improve operations and generate higher revenue.”
Please disclose the source of your assertion that you pioneer and deliver transformative AI-enabled solutions and that you are a leader in this space.
Here and in the Business section, revise to clarify the development status of the Company’s conversational AI technology. Describe the scope of the current capabilities of the Company’s conversational AI tech platform that supports or will support the Company’s products and/or services, and clearly distinguish current technological capabilities from future aspirational capabilities. Describe any material steps that will need to be taken to fully develop the Company’s platform and technologies.
Please revise the bullet points on page XX of the proxy statement to clarify, if true, that these are not yet products or services the company provides, and are instead areas of research or are aspirational.
AI-related disclosures about a company’s business should be specific and balanced, as well as identify material operational, legal, competitive, and similar risks. Companies should avoid generic boilerplate language when providing these disclosures. Notably, approximately 61% of the SEC’s comments we reviewed requested that companies that have disclosed AI-related initiatives, projects, or technologies clarify how the AI is or is intended to be used in those initiatives, projects, or technologies and any attendant risks. Examples of such SEC comments include:
We further note the disclosure that broad-based AI adoption is in its early phases and that AI-adoption is likely to continue and may accelerate. Please revise your business section to provide a more balanced discussion of AI. Include, without limitation, a discussion of the potential limitations, obstacles, and uncertainties associated with AI adoption, use, and commercialization.
We note your disclosure that the product is designed with “extensive AI skill sets and machine learning capabilities.” Please revise to explain how your product uses artificial intelligence and machine learning.
Please revise to clarify the liability that you assume, if any, if your AI technology incorrectly evaluates the creditworthiness, likelihood of default, or credit asset pricing for your Partners’ customers.
Your disclosure suggests your AI models use a diverse set of inputs including news stories and sentiment analysis, market indicators, and company fundamentals, among others to inform the decision-making process. Clarify how these inputs are structured and utilized and how they are presented to customers to provide actionable data and how that data can be used by your customers.
Include an appropriate risk factor addressing the material risks to your business, operations, and financial condition in connection with the use and development of products using generative artificial intelligence technology, as appropriate.
Additionally, to the extent material, include a cross reference to risk factor disclosure discussing the operational, legal, and competitive risks to using AI.
You reference that your AI platform and data it uses may subject you to additional risks under HIPPA and other health privacy laws. Please clarify whether you are referring specifically to the use of AI in the operation of your platform and whether such use may expose you to risks of non-compliance with such health privacy laws. For example, whether the AI functionality may cause your controls and procedures related to privacy protection to be avoided or overwritten without proper authorization or oversight.
Companies should define what they mean by AI and related terms, including how these terms are used in their specific business context. Our survey showed that 17% of the SEC’s comments we reviewed addressed the use of AI-related terminology and definitions. Examples of such SEC comments include:
Given the nature of your business, please consider including definitions of “AI,” “generative AI,” “deep learning,” “large language models,” “neural networks,” and any other industry-specific terminology. Please provide your definition of “artificial intelligence” in the context of your business and enhance your disclosure to provide an update on the stage of product development for your Platform. Please explain how your software is properly characterized as AI or machine learning, rather than as an algorithm. Please revise to more clearly explain what “dark” and “unstructured” data are, as well as how [your] AI [program] differs from artificial intelligence more generally.
Beyond these categories, 34% of the SEC’s comments we reviewed addressed other AI-related issues, such as IP, the collection and use of data implicated in AI applications, the involvement of third parties, how the AI was developed, validation of models, and disclosure inconsistencies. Examples of such SEC comments include:
With respect to artificial intelligence, please indicate if your algorithms are proprietary or open source, and if open source, whether the AI is governed by open source licenses, and update your risk factors to reflect the relevant risks.
Please revise to clarify how you obtain the datasets you utilize. For example, it is unclear whether you use only publicly available data or if you use APIs or have access to third-party proprietary datasets, and the costs and limitations associated with each.
Please also describe the types and quantities of data you used, or will use, to train the algorithms, where you will collect this data, and provide a timeline of when you expect to commercialize these products.
In addition, please disclose the circumstances under which collected data is shared or sold to third parties.
To the extent the AI model was not developed by you, clarify if it is an off-the-shelf AI program or a model provided by third parties or if they were pre-selected algorithms, AI models, or chatbots.
Please expand your disclosure to explain how you developed and validated your artificial intelligence and machine learning algorithm(s). Please describe the frequency with which validation occurs and how long the model has been commercialized.
You claim that your software does not rely on large historical data sets, yet in the following sentence, you claim that your advanced machine learning/AI algorithms combine streaming data with historical information. Please reconcile.
The 56 companies receiving these comments were well distributed across various industries, including technology, biopharmaceuticals and healthcare, real estate, retail, and financial services. For both domestic and foreign companies, the SEC has commented on disclosures in traditional registered offerings (via Forms S-1, S-3, S-4, F-1, and F-4), in Reg A and Reg A+ offerings, in filings for the registration of securities not in connection with an offering (via Form 10-12G), as well as in proxy materials and annual reports. Accordingly, companies across all sectors should consider the SEC’s guidance and comments that have been issued to date as they prepare various disclosures and make public statements, including in press releases and during earnings calls. By incorporating these insights into disclosure practices, companies may avoid additional review or enforcement actions by enhancing the accuracy and relevance of their AI disclosures.
[1] AI washing is generally deemed to occur when a company overstates its AI capabilities or use of AI to mislead the market.
[2] Some SEC comments addressed multiple categories simultaneously, resulting in their inclusion in more than one category. Consequently, the percentages mentioned add up to more than 100%.
[3] The SEC comments presented in this article have been genericized to highlight main themes, leaving out specific details about individual companies and products.