AI supervision and the 2026 exam priorities.

What the 2026 priorities actually say about AI.

The Division of Examinations publishes annual priorities that signal where its examiners will focus. The fiscal year 2026 priorities (released November 2025) include a section on emerging financial technology that names automated investment tools, AI technologies, and trading algorithms as areas of focus. The document is a staff statement; by its own terms it has “no legal force or effect” and “creates no new or additional obligations.” It describes where examiner attention is heading, not a new rule.

Two points in that section stand out for a firm that uses AI:

• Supervision of AI use.The Division said it will assess “whether firms have implemented adequate policies and procedures to monitor and/or supervise their use of AI technologies,” and will look for “controls to confirm that advice or recommendations resulting from automated tools are consistent with regulatory obligations to investors, including retail and older investors.”
• Accuracy of AI claims.The Division said it will “review for accuracy registrant representations regarding their AI capabilities.” In plain terms, if a firm describes what its AI does, the accuracy of that description is something examiners have said they may review. This is the area commonly described as “AI-washing.”

What that signals in practice.

Read together, the focus is on two questions a firm using AI should be able to answer: how is the AI overseen, and is the firm's description of what the AI does accurate. Neither question is about whether AI may be used. Both are about whether its use is supervised and described honestly. On the SEC's framing, the priorities apply an existing framework to AI rather than creating a new rule: the fiduciary duty, the compliance program rule, the marketing rule, and the books-and-records requirements that firms and their counsel already work with. That is why this reads as a direction of travel rather than a current mandate, though what any of it requires is a determination for the firm and its counsel.

What a documented, supervised review looks like.

One practical response to “how is the AI overseen” is a contemporaneous record. When AI is used somewhere in a decision, a record made at the time can show what was surfaced, that a licensed person reviewed it, who approved it, and why. Many firms find the harder position is the reverse: AI was used, and later there is no record of what was reviewed or by whom. Firms meet this in many ways: contemporaneous notes, investment-committee minutes, and existing supervisory records all do work here.

One such record is what Teranode produces. The Council runs several independent reasoning roles against a question and returns a signed, timestamped decision record: the decision, the risks that were surfaced, a confidence figure (a measure of how strongly the roles converged, computed from a published, deterministic formula, not a probability of being correct), and the preserved dissent attributed to the role that raised it. Each record carries a SHA-256 fingerprint of its contents, so two copies can be compared for an exact match. It is a signed, timestamped decision record, not a tamper-proof seal, and the adviser owns the decision. The formula and method are on /methodology.

What Teranode does not do.

Teranode is decision-support software. It does not make recommendations to your clients, it does not certify compliance, and it does not make any firm ready for an examination. It produces a structured record of a supervised reasoning process that your firm's licensed people review and sign. Whether to keep such records, what they should contain, and whether any disclosure is required are determinations for your firm and its counsel. Teranode does not provide regulatory determinations.