teranodeSEC AI Brief · archive

Past issues

Dated summaries of SEC, AI, and adviser-governance developments, with links to the primary sources.

July 21, 2026
FINRA's Rule 2210 proposal would make AI supervision more risk-based, not less documented
FINRA's July 9 Regulatory Notice 26-14 requests comment on a proposed modernization of Rule 2210, not a current rule change. Comments close September 11. Today, an appropriately qualified principal generally must approve retail communications before use, subject to specified exceptions. The proposal would instead require written procedures, tailored to the member's business, size and structure, that determine which categories still require principal pre-use approval. The procedures would have to meet Rule 2210's unchanged content standards. If not every communication is approved before distribution, they should include training, documentation, surveillance and follow-up; firms would maintain evidence that the procedures were implemented. Risk factors include product complexity, preparer qualifications, recommendations or promotions, audience, distribution method and performance information. AI is in scope but not exempt. FINRA says members remain responsible for communications whether produced by a person or AI, and GenAI tools may participate in a supervisory system when vetted, tested and monitored. Higher-risk communications could receive more rigorous pre-use review while lower-risk communications follow a streamlined path. Current Rule 2210 remains in force while FINRA considers comments and any later rule filing.
Why it matters for advisers
For FINRA members and dual registrants, the operational question is not whether supervision disappears. It is how the firm classifies communication risk, documents the review path it selected and demonstrates that its procedures operated. A written policy describes the system; it does not show what happened on a particular communication. A run-level record can support that file, but it is only one component of the firm's governance, training, surveillance and escalation program. RIA-only firms are not governed by this FINRA proposal. For them, the notice is an industry signal rather than an applicable standard, and their federal or state adviser obligations remain separate.
Teranode read
RN 26-14 draws a boundary between enterprise tool governance and evidence created during an individual review. A firm must vet, test and monitor a GenAI tool at the program level; no single review record can establish those controls. At the run level, a record can show source metadata, findings, human dispositions, final decision, review vendors and whether the source vendor was excluded. Structural separation matters because a model or vendor evaluating its own output may reproduce the same blind spots or incentives, ours included. Cross-vendor review, however, is not third-party assurance, a compliance certification or a safe harbor. Each firm decides its procedures, which categories require principal pre-use approval, what it retains and when it escalates, with its own counsel and compliance professionals. Teranode's claim is narrower: a record can make review structure and human disposition inspectable while the firm remains responsible for supervision and the final determination.
Primary sources
June 5, 2026
SEC signals the AI question is moving from “can advisers use it?” to “how is it supervised?”
Across recent SEC speeches, the FY2026 examination priorities, and a 2025 rulemaking withdrawal, a consistent picture emerges for investment advisers. The Division of Investment Management has named artificial intelligence one of its priorities and frames its posture as enabling and supporting AI while regulating it thoughtfully; Director Brian Daly has described adviser adoption as still uneven and tentative, with liability fear as the main blocker and the human role shifting toward supervision, while inviting industry engagement through avenues such as no-action relief and staff guidance and noting that the books-and-records regime is dated. The Division of Examinations, in its FY2026 priorities, names automated investment tools, AI technologies, and trading algorithms as focus areas, stating it will review the accuracy of registrants’ representations about their AI capabilities and assess whether firms have adequate policies and procedures to monitor and supervise their use of AI. The Commission’s withdrawal of the 2023 predictive-data-analytics conflicts proposal removed that specific framework, leaving the underlying fiduciary and conflicts standards in place. And Chairman Paul Atkins, speaking at the SCSP AI+ Expo, reiterated a principles-based stance: foundational obligations do not change, firms remain responsible for the outcomes of the tools they deploy and for informing investors how those tools are used, and scale is what is genuinely new about AI.
Why it matters for advisers
The center of gravity for advisers is moving from whether AI use is permitted to whether it is governed. Examiners have said in writing that they will test AI-capability representations for accuracy and look for policies and procedures that monitor and supervise AI use, so an adviser deploying AI should expect questions about supervision, disclosure accuracy, and how AI-influenced recommendations square with fiduciary duty and the firm’s conflicts obligations. The withdrawn predictive-data-analytics proposal does not relax anything: the Advisers Act fiduciary standard and existing conflicts, marketing, and books-and-records rules still apply to AI-assisted activity, and any future AI-specific rule would require a fresh proposal.
Teranode read
The throughline across these sources is supervision and documentation: regulators are signaling that AI use is acceptable when a firm can show how it is overseen, how its capabilities are described, and how conflicts and fiduciary duties are addressed. For advisory firms, that raises the practical value of a documented, supervised decision process, a defensible record of how a recommendation was reached, what was considered, and who reviewed it. This is an editorial observation about where examination attention is heading, not legal advice and not a claim that any particular tool or record satisfies a specific SEC rule; firms should work that question through their own counsel and compliance professionals.
Primary sources

Stay current

Signal on SEC, AI, and adviser-governance developments. Built for investment advisers tracking supervision, records, fiduciary process, and AI governance.

Join the SEC AI Brief

Past issues

FINRA's Rule 2210 proposal would make AI supervision more risk-based, not less documented

SEC signals the AI question is moving from “can advisers use it?” to “how is it supervised?”