RIA Disclosures Reveal AI as a Growing Risk Factor Rather Than a Standard Capability
For registered investment advisors (RIAs), the rapid integration of artificial intelligence into financial workflows has created a complex regulatory and communication challenge. According to a new peer-reviewed study published October 14, 2026, in Frontiers in Artificial Intelligence, RIAs are significantly more likely to categorize AI as a potential risk within their Form ADV Part 2A brochures than they are to frame it as a core service capability.
The study, which analyzed 388 Form ADV brochures, highlights a growing discrepancy between the technological adoption of these tools and the language used to describe them to clients and regulators. As the Securities and Exchange Commission (SEC) formalizes its scrutiny of AI, the findings suggest that the investment industry is currently navigating a period of significant uncertainty regarding how to properly document the use of emerging technologies.
Inside the 388-Brochure Sample
The research was conducted by Samir Chincholikar and Robin Chawla, independent researchers based in New York City. To arrive at their findings, the pair drew a stratified random sample of 400 SEC-registered investment advisors from a total population of 16,223 firms active as of the end of 2024. The sample was carefully segmented by advisor type—distinguishing between private-fund managers and retail-focused wealth managers—and further divided by assets under management (AUM) quartiles. After accounting for available documentation, the final dataset consisted of 388 usable brochures.
To process this volume of legal documentation, the authors utilized a large language model, specifically GPT-4o, to categorize each brochure under five distinct labels: AI in the investment process, AI in operations or client service, AI as a disclosed risk, explicit non-use, and the identification of specific AI vendors. Crucially, the researchers did not treat the AI’s output as infallible. Instead, they used it as a measurement instrument, subjecting the model’s findings to a blinded recoding process by Anthropic’s Claude, followed by rigorous human review to ensure accuracy.
The study revealed that 23.7% of the brochures disclosed the use of AI (or 22.7% when adjusted for survey weighting), while 27.6% contained language specifically addressing AI-related risks. The researchers noted that the mere presence of the term "artificial intelligence" in a brochure is often misleading if viewed in isolation. As Chincholikar explained, the term is frequently invoked to describe the dangers or risks associated with automated investment processes rather than the firm’s active deployment of such technology to enhance client portfolios.
The composite measurement tool developed for the study achieved a precision score of 0.76 and a sensitivity score of 0.81 when compared against human coding. This means that roughly three out of four brochures flagged by the AI were confirmed by human reviewers, and the automated tool captured about four out of five instances identified by human experts. The authors noted that a simple keyword search—a method often used by casual observers—would have missed or mislabeled approximately 25% of the relevant disclosures, underscoring the necessity of context-aware analysis.
Within the sample, 77 brochures carried both risk language and evidence of AI use. However, 30 brochures contained risk language without any mention of active use, while 15 firms disclosed use without explicitly addressing the risks. A large majority, 266 firms, made no mention of AI in either capacity. The 3.9-percentage-point gap between risk language and use was found to be statistically significant, a margin that widened to 5.2 points when references to third-party vendors were stripped away. Chincholikar suggested that this trend is consistent with the nature of the Form ADV, which is designed to be a repository for material risks rather than a marketing document for a firm’s technological prowess.
Size and Business Model Drive the Gap
The research uncovered a clear correlation between firm size and the propensity to disclose AI activity. Disclosure rates climbed significantly as AUM increased, rising from 10.4% in the smallest quartile to 43.8% in the largest. This trend held true across different business models as well. For private-fund advisors, disclosure rates increased from 12% to 56% across the AUM spectrum, while wealth and retail advisors saw an increase from 8.7% to 30.4%.
Chincholikar attributed this disparity largely to compliance infrastructure. Large firms typically possess the resources to maintain expansive compliance teams, which provides them with a deeper, more granular understanding of their own internal processes. Conversely, smaller firms may be using similar tools but lack the specialized manpower to document those uses in the highly specific language required for regulatory filings. However, Chawla cautioned against interpreting these numbers as a sign of institutional dishonesty. The researchers emphasized that there is no statistical evidence to suggest that smaller firms are intentionally obfuscating their use of AI; rather, they may simply lack the institutional capacity to formalize their disclosures.
Navigating the "Benchmark Divergence"
One of the most complex aspects of the study is reconciling these findings with broader industry surveys. For instance, the 2026 RIA and AI Research Study from Charles Schwab found that 63% of 533 Schwab-custodied RIAs were using AI in some capacity by October 2025. The roughly 39-percentage-point gap between the Schwab survey and the researchers’ brochure analysis is characterized by the authors not as a nondisclosure crisis, but as a "benchmark divergence."
Many of the AI tools used by RIAs—such as internal administrative assistants, email drafting software, or experimental data processing tools—may be considered operational background noise rather than material information that requires disclosure in a regulatory document. The authors also investigated whether websites offered a more transparent view, but found similar results. Among the 283 firms that could be observed both on their websites and in their brochures, only 5.7% disclosed AI use on their public-facing websites, compared to 23.0% in their brochures. However, given that brochures typically contain significantly more text than websites, the researchers concluded that the difference is largely a function of document length rather than a lack of candor.
The SEC’s Evolving Regulatory Stance
The study serves as a timely warning for firms as the SEC ramps up its oversight of the technology. The researchers noted that relying on a single, aggregated number to describe AI adoption in the industry is dangerous, as the results can vary wildly—from 9.9% to 31.7%—depending on how one defines "use" and how the data is weighted. This ambiguity presents a genuine compliance risk.
The SEC has already signaled that AI is a top priority. In March 2024, the agency took enforcement action against two firms, Delphia and Global Predictions, for allegedly overstating their AI capabilities in their brochures. Chincholikar and Chawla used the language from these SEC orders to build a "screening" tool to test the 388 brochures in their study. The results were largely benign; the screen flagged only five brochures, and of those, two were false positives, while the remaining three contained legitimate, context-qualified references to AI. None of the brochures were found to contain the type of unsubstantiated, "AI-washed" promotional claims that drew the SEC’s ire in 2024.
As of October 23, 2026, the SEC’s fiscal year 2027 examination priorities have officially added AI-related cyber risk to their list of concerns. Furthermore, following an executive order in September 2026, the agency has begun referring to the technology as "super intelligence" in its regulatory communications, signaling a shift toward more sophisticated oversight.
For advisors looking to stay ahead of this evolving landscape, the researchers offer a practical, firm-wide recommendation: maintain a comprehensive, living inventory of all AI tools used within the firm. By clearly delineating where each tool is applied and the specific risks it may introduce, advisors can move beyond the current environment of uncertainty and ensure their disclosures align with both their actual operations and the heightened expectations of their regulators.
| Source | : | Finance Poetry |
| Author | : | rifanmuazin |
| Editor | : | Finance Poetry Editorial Team |
