Legal and compliance departments at private fund managers find themselves at an inflection point. On one hand, how do they support their firm’s artificial intelligence (now called super intelligence) adoption journey? On the other, how do they manage this transformation, as they will be the ones narrating it to the Securities and Exchange Commission (SEC)?
Super intelligence (SI) seemingly has the power to change everything, but not the SEC or the law. The SEC’s position persists that, regardless of technology, a manager’s obligations to clients remain unchanged. The SEC staff reiterated this in the SEC’s 2027 Examination Priorities:
“As [super intelligence] capabilities grow more sophisticated and embedded in daily business processes, the Division is intensifying its efforts to understand how the technologies influence outcomes and whether firms continue to comply with existing, technology-neutral obligations.”
A manager’s use of SI is a business decision. In many ways, it is akin to selecting a new vendor, procuring a new tool, or automating an existing workflow: something the firm used to do one way is now evolving. Nothing about that changes the existing laws that apply or the SEC’s expectation that a manager acts in the best interests of its clients.
With this in mind, in the 2027 Examination Priorities the SEC staff says that, in assessing a manager’s use of SI, it will consider whether the manager’s “(1) representations are fair and accurate; (2) operations and controls are consistent with disclosures made to investors; and (3) algorithms lead to advice or recommendations consistent with investors’ investment profiles or stated strategies.” In many ways, these are the basics: accurate disclosures and operations, controls and trading consistent with them. Although easier said than done, the SEC’s confirmation that the basics still apply gives managers a familiar framework in which to make the most of SI in their firms.
Element #1: Fair and Accurate Representations
Each element above circles back to disclosures. Disclosures include a firm’s Form ADV, offering documents, pitchbooks, and due diligence questionnaires (DDQs). A disclosure-based approach, familiar territory to any manager, can provide flexibility and grace in SI implementation.
Form ADV. Firms should consider whether their Form brochure accurately describes the “methods of analysis” and “investment strategies” the firm uses to formulate investment advice. In answering this question, as well as other aspects of the ADV, firms will need to self-assess how they use SI. An accurate self-assessment may be the most challenging part of the process. Additionally, because the federal securities laws do not define SI, how a firm describes the kind of SI it uses is itself a disclosure decision. As SI usage begins to mature, managers should expect examiners to know the difference between machine learning and generative tools and to expect the firm to know which ones it uses.
AI Governance. SEC examinations of SI usage increasingly resemble examinations of any significant business function: the staff may ask who at the firm is accountable, what the firm has approved, and what processes the firm follows. At a minimum, AI governance includes an inventory of SI tools a firm uses and what each is permitted to do, named senior personnel responsible for oversight of SI, including determining who may use which tools and connect which data. Although the SEC has not delivered specific guidance on SI, various international organizations have, and their guidance touches on the same elements (senior accountability, pre- and post-deployment testing, qualified personnel, data quality controls, third-party oversight, and disclosure of SI reliance and risks).
AI Washing. The SEC’s only SI-related enforcement actions against advisers to date involved firms that overstated their SI usage. Exaggeration remains a clear and somewhat obvious risk. However, understating a firm’s SI usage is a more likely and a more subtle risk. A robust self-assessment and formal AI governance can help a firm understand and therefore accurately describe its SI usage to clients.
Element #2: Operations and Controls Consistent with Disclosures
The second element of the 2027 Examination Priorities focuses on whether the manager’s operations and controls match its disclosures. Here are a few key areas for managers to consider:
Data. Modern fund managers run a disciplined process for onboarding market and alternative data. Specialized departments conduct vendor diligence, implement contractual representations, and jointly approve datasets before they’re deployed into production. With the toggle of a button, however, SI applications can collapse these processes entirely. SI tools can easily pull unauthorized data directly into the firm. Employees can also load data files directly into SI tools. Agentic SI tools go a step further: given a research task, an agent can retrieve outside data on its own, with no person making a deliberate ingestion decision at all.
Using SI in research or elsewhere in the investment process is plainly a meaningful step. However, connecting SI to a data source, whether internal or external, is itself a discrete step, and one that has different implications. Data should not reach an SI tool unless it has passed through a deliberate intake process, such as the one a firm applies to any purchased dataset. Firms should also confirm that written policies, including on material non-public information (MNPI), are updated to describe how SI can affect the data intake process.
Information Barriers and MNPI. Managers maintain electronic and physical information barriers to segregate data within their firm, for example between trading functions and other parts of the firm. Electronic information barriers can include restricted access to certain folders or tools. Information barriers not only wall off external data, as described in the prior bullet, but internal data as well. Firms should ensure SI usage does not result in employees accessing walled-off data. The risk also manifests when a firm trains its own SI on internal data. Models can become carriers. And employees who can’t access restricted data may still receive its contents — or inferences drawn from them — by using shared SI models.
Recordkeeping. Advisers are currently tasked with the formidable project of ensuring their books and records systems and policies capture SI output. Some processes, such as standard email review, may not change much. Other processes, however, raise novel questions: when is output from SI a book or record? Additionally, SI has the potential to create a drastically larger corpus of internal “thinking.” In the past, an internal slide deck may have been created with the employee iterating on the file, saving, and overwriting versions, etc. Now, employees may iterate on the slide deck via an SI tool, which will record each iteration and question and piece of feedback — in a sense, prior versions of the file — creating sensitive and auditable firm records. Firms will need to both update their processes to cover AI-generated books and records and evaluate their use of AI from a broader litigation and enforcement perspective.1
Element #3: Advice Consistent with Stated Strategies
The third element of the 2027 Examination Priorities reaches the investment decision itself: is the firm still trading within each fund’s stated strategy and investment objectives?
Model Governance. Systematic managers have long run formal model-approval processes. A model or modification is submitted, reviewed, tested, and approved before it is deployed in production. Like many things in fund compliance and operations, this is a process. It captures who submitted the change, who reviewed it, what testing occurred, when it went live, and related supervision. Given the power of SI, the temptation or possibility exists to shortcut or once again collapse these processes. SI can likely perform most or all of these functions: it can ideate, generate, review, and test. Firms should remain vigilant about the difference between using SI within a step versus letting SI absorb a process itself. SI can usefully assist — drafting submissions, generating test cases, and summarizing results — so long as the separation between steps, and with it the process and explainability that are hallmarks of a compliant program, remains.
New Capabilities and Strategy Expansion. SI can unlock capabilities a firm did not previously have, such as a greater ability to structure data or to model a new instrument. Each capability expansion raises the question of whether it is within the fund’s stated strategy or whether the strategy has expanded. In the priorities, the SEC staff separately flagged business drift — including advice regarding new asset classes or new services — as an examination focus.
The Road Ahead
Managers should not wait for further rulemaking to resolve these questions. The SEC withdrew its proposed AI conflicts rule in June 2025 and stated that it does not intend to adopt it. Private fund advisers should separately note the priorities’ attention to fund-level and investment-level controls, including fee and expense calculations, and the enhanced SEC–CFTC memorandum of understanding contemplating coordinated examinations of dual registrants. The encouraging message in the priorities is that the SEC staff has not asked managers to build anything new. Disclosure review, data intake and information barriers, approvals, and processes — these are disciplines managers already run. The challenge for firms is likely more of an internal resourcing question and vigilance question, not a legal one.
For questions about how these developments may affect your business, please contact the authors.
1That evaluation should also consider attorney-client privilege. A conversation with an SI tool is not a conversation with counsel. Candid internal “thinking” these tools record — drafts, questions, and analysis — can be discoverable in litigation or by a regulator.
You are switching to the United States
This selection will switch the website from presenting information primarily about the United Kingdom to information about the United States. If you would like to switch back, you may use location selection options at the top of the page.
Contact
Although we would like to hear from you, we cannot represent you until we know that doing so will not create a conflict of interest. Also, we cannot treat unsolicited information as confidential. Accordingly, please do not send us any information about any legal matter until we authorize you to do so. To initiate a possible representation, please call one of our lawyers or staff members.
By clicking the “ACCEPT” button, you agree that we may review any information you transmit to us. You recognize that, even if you submit information that you consider confidential in an effort to retain us, our review of that information will not create an obligation on us to keep it confidential and will not preclude us from representing another client directly adverse to you, even in a matter where that information could and will be used against you.
Please click the “ACCEPT” button if you understand and accept the foregoing statement and wish to proceed.