When AI Meets the Examiner

1 hour ago

1 min read

DeepVest Research. (2026, November). When AI meets the examiner. DeepVest. https://www.deepvest.ai/articles/when-ai-meets-the-examiner

The SEC's FY2027 Examination Priorities, released October 23, 2026, direct examiners to check whether firms' claims about their AI tools are accurate and whether firms are monitoring how they use them.
Treasury's Financial Services AI Risk Management Framework, released in February, sets out roughly 230 control objectives for managing AI risk, including risk from outside vendors.
The article states that AI adoption among advisors has outpaced evaluation, with tools now used to summarize meetings, draft client emails, research markets, and run portfolio calculations.
The paper reduces both regulatory sources to three questions for any AI tool: whether it can explain its outputs, whether its results can be verified, and whether the vendor commits in writing to client data, security, and audit-trail terms.
For verification, the article recommends firms keep pre-adoption test results in the compliance file, including the questions asked, answers received across repeat runs, and the reference values used.
In an earlier DeepVest study of six AI systems running ten calculations five times each, Perplexity's answers to one Sharpe ratio question varied with a 93.7% coefficient of variation, and ChatGPT's correlation answers varied by 43.7%.
The article notes that fiduciary duty does not shift when an AI tool helps produce advice, and that the evidence a firm needs scales with how close a tool's output gets to a client's portfolio.

Knote: We continue to believe that AI Governance will be one of the big themes of next year.