At a glance
- The agency's 2026 exam priorities cover the training and security controls firms use to identify and limit new AI risks.
- The main client-data risk is employees who build their own AI tools, according to the report.
- Examiners may ask whether written policies cover employee-built agents and which AI accounts staff may use, rather than only the third-party software RIAs vet closely for client-data risk.
The SEC is examining how advisory firms train and supervise employees on technology, including AI, InvestmentNews reported on Nov. 16. The agency's 2026 exam priorities cover the training and security controls firms use to identify and limit new AI risks.
The main client-data risk is employees who build their own AI tools, according to the report. Free versions of ChatGPT or Claude generally use entered information for model training by default, and an employee-created agent could bypass firm safeguards even when no client data is typed in. RIAs must protect that client data under SEC Regulation S-P.
Examiners may ask whether written policies cover employee-built agents and which AI accounts staff may use, rather than only the third-party software RIAs vet closely for client-data risk.
Reported by the RIA desk from InvestmentNews. Published . Corrections