What the SEC's 2026 Exam Priorities Mean for AI Use

The two changes that matter

First: AI washing is named. Second: AI oversight is now tested in effectively all examinations rather than as a specialist topic.

Per Goodwin's summary of the 2026 priorities for registered investment advisers, examiners will test the policies and procedures firms have for supervising AI, and firms are required to ensure marketing, Form ADV disclosures and client communications accurately describe AI's "extent, nature, and limitations."

What AI washing looks like in practice

It is rarely a deliberate lie. It is usually a marketing sentence that outran the product.

The common patterns: describing a feature in development as though it ships today; describing a model's output as analysis when it is summarisation; claiming a capability the vendor supplies without disclosing the vendor's role; and repeating a vendor's own accuracy claim without evidence for it.

The exposure is specific to marketing surfaces a compliance function may not routinely review — the website, the pitch deck, the conference booth, the sales email. Those are client communications when they reach clients or prospects.

Form ADV disclosure

Disclosure must describe what the AI does, what it does not do, and what its limitations are.

Two failure modes are common. Silence — using AI materially in the advisory process without describing it. And overstatement — describing an AI capability in terms that make it sound more autonomous or more accurate than it is. The 2026 priorities make both examinable.

Supervision is the tested control

Examiners test policies and procedures for supervising AI, which means the artefact under examination is your documentation, not your technology.

This is why the survey data matters so much in 2026. Against 80% adoption of AI tools, 48% of firms had human-in-the-loop oversight policies and 37% had procedures to test or validate outputs before client delivery, per the compliance testing survey reported by InvestmentNews. A firm with sophisticated AI and no written supervision procedure is worse positioned at examination than a firm with modest AI and a documented one.

What to do before the next cycle

• Inventory every AI system in use, including AI inside tools adopted for another purpose.
• Read your own marketing as an examiner would, and substantiate or remove each capability claim.
• Review Form ADV language against what the systems actually do today.
• Document the human review step and make sure it leaves a record, not just a rule.
• Record the limitations you know about. A disclosed limitation is not a weakness at examination; an undisclosed one is.

A note on timing

This page describes the 2026 priorities and will be revised when the 2027 priorities publish. Examination priorities are directional rather than exhaustive — a topic's absence from the list does not place it out of scope.