The Growing Use of AI in Investment Advice
Artificial intelligence and machine learning technologies are rapidly transforming how investment advice is delivered, portfolio management is conducted, and trading decisions are executed. From robo-advisors serving retail investors to sophisticated algorithmic trading systems used by institutional managers, AI-powered tools have become integral to modern asset management.
While these technologies offer significant potential benefits including improved efficiency, data-driven insights, personalized recommendations, and scalable service delivery, they also introduce novel compliance risks and regulatory challenges. The SEC has made clear that investment advisers using AI tools remain fully responsible for the advice provided and must ensure that AI systems operate consistently with fiduciary obligations and regulatory requirements.
In 2024 and early 2025, the SEC issued guidance clarifying how existing regulatory frameworks apply to AI-powered advisory services. This guidance addresses key areas including algorithm development and testing, ongoing monitoring and validation, disclosure obligations to clients, conflicts of interest management, and books and records requirements for AI systems.
SEC Guidance on AI Use in Advisory Services
The SEC's approach to AI regulation emphasizes that advisers using these technologies must maintain effective oversight and control over algorithmic systems. The Commission has stated that reliance on AI tools does not diminish an adviser's fiduciary duty to act in clients' best interests or excuse failures to provide advice suitable for client circumstances.
Key regulatory expectations include thorough testing and validation of AI algorithms before deployment, ongoing monitoring of algorithm performance and outputs, procedures to identify and correct algorithm errors or biases, human oversight of critical investment decisions, and robust documentation of algorithm logic, parameters, and decision processes.
The guidance also addresses disclosure requirements, emphasizing that advisers must provide clients with clear, accurate information about how AI tools are used in the advisory process. This includes explaining the role of algorithms in generating recommendations, describing limitations of AI systems, disclosing potential conflicts of interest, and clarifying the extent of human involvement in advice delivery.
Additionally, the SEC has focused on preventing what it terms "AI washing," the practice of overstating or misrepresenting AI capabilities in marketing materials. Advisers making claims about AI-driven performance, advanced analytics, or machine learning sophistication must substantiate these representations with evidence and ensure marketing does not mislead investors about technology capabilities.
Building an AI Governance Framework
Implementing effective AI governance requires a comprehensive framework addressing technology development, deployment, monitoring, and risk management. This framework should begin with clear policies establishing the adviser's approach to AI use, approval processes for new AI tools, roles and responsibilities for AI oversight, testing and validation standards, and escalation procedures for AI-related issues or failures.
The governance structure should include senior management oversight, technical expertise to evaluate AI systems, compliance involvement in algorithm review, independent testing and validation, and clear accountability for AI system performance. Many advisers establish AI governance committees or working groups to coordinate oversight across technology, compliance, risk management, and investment teams.
Documentation is critical for demonstrating effective AI governance. Advisers should maintain detailed records of algorithm design specifications, training data sources and characteristics, model testing and validation results, ongoing performance metrics and monitoring, identified issues and remediation actions, and decision-making processes for algorithm updates or modifications.
Algorithm Testing and Validation
Thorough testing and validation are essential before deploying AI algorithms in client-facing applications. Testing should evaluate algorithm accuracy and reliability, consistency with stated investment objectives, performance across different market conditions, potential biases in recommendations, and alignment with adviser's fiduciary obligations.
Advisers should conduct both back-testing using historical data and forward-testing in controlled environments before full deployment. Testing protocols should include stress testing under adverse scenarios, sensitivity analysis to understand key drivers, comparison against benchmark strategies or human-generated advice, and evaluation of algorithm behavior at portfolio boundaries or edge cases.
Validation should be performed by personnel independent from algorithm development to provide objective assessment. This independent review should verify that algorithms operate as intended, produce appropriate outputs, incorporate relevant inputs and constraints, and align with documented design specifications and investment principles.
Ongoing Monitoring and Model Risk Management
AI algorithms require continuous monitoring to detect performance degradation, unintended behaviors, or emerging risks. Monitoring programs should track algorithm outputs and recommendations, error rates and failed transactions, deviations from expected performance, client complaints or concerns, and unusual patterns in generated advice.
Model risk management is particularly important as market conditions, client populations, and regulatory requirements evolve. Advisers should establish procedures for periodic model review and revalidation, assessment of changing market dynamics, evaluation of new data sources or inputs, consideration of regulatory or business changes, and determination of whether model updates or replacements are needed.
When algorithm issues are identified, advisers must have clear escalation and remediation processes. This includes immediate assessment of client impact, temporary suspension of affected algorithms if necessary, root cause analysis to understand failures, corrective action implementation, and communication to affected clients when appropriate.
Disclosure and Client Communication
Transparent disclosure about AI use is fundamental to meeting fiduciary obligations. Form ADV disclosures should describe how AI tools are used in the advisory process, the types of algorithms employed, the role of human oversight, limitations or risks of AI-driven advice, and how clients can request human review of algorithm-generated recommendations.
Client communications should explain AI functionality in accessible language appropriate for the target client base. Overly technical descriptions may confuse rather than inform, while oversimplified explanations may fail to provide meaningful disclosure. Advisers should strive for balance, conveying material information about AI use without overwhelming clients with technical details.
For advisers using AI in marketing, all performance claims or capabilities statements must be accurate and substantiated. Claims about AI sophistication, superior returns, or advanced analytics require supporting evidence and should not mislead investors about actual technology capabilities or expected outcomes.
Addressing Bias and Fairness Concerns
AI algorithms can inadvertently perpetuate or amplify biases present in training data or embedded in model design. Investment advisers must proactively identify and address potential biases to ensure fair treatment of all clients and compliance with applicable laws including anti-discrimination requirements.
Bias testing should evaluate whether algorithms produce systematically different outcomes for client subgroups, whether certain populations receive less favorable advice, whether data sources contain historical biases, and whether algorithm design incorporates inappropriate proxies or correlations. When biases are identified, advisers should implement corrective measures such as data augmentation, algorithm retraining, or additional oversight controls.
Key Takeaways for RIAs
- The SEC expects advisers using AI to maintain effective oversight, control, and accountability for algorithm-driven advice.
- Comprehensive AI governance frameworks should address development, testing, deployment, monitoring, and risk management.
- Thorough testing and independent validation are required before deploying AI algorithms in client-facing applications.
- Ongoing monitoring and model risk management help identify performance issues, biases, and emerging risks.
- Clear disclosure about AI use, capabilities, limitations, and human oversight is essential for fiduciary compliance.
- Advisers must substantiate marketing claims about AI sophistication and avoid misleading investors about technology capabilities.
- Proactive identification and remediation of algorithm biases ensures fair treatment and regulatory compliance.