The Problem: AI Descriptions Without a Trace
A machine is describing your business to prospects several times a day, and there is no record of any of it. This was the genesis of Attest, an agent designed to audit what other AI models say about SEC-registered investment advisers. The core issue is the increasing reliance on AI to interact with potential clients, coupled with a complete lack of accountability or auditability for the information disseminated. This gap becomes critical when considering regulatory scrutiny and the potential for misinformation or misrepresentation.
The SEC's 2026 examination priorities explicitly mention AI, particularly concerning the fairness and accuracy of a firm's claims about its own AI. While regulators are looking inward at how firms present their AI capabilities, the broader legal landscape regarding liability for what third-party AI models say about an investment adviser remains largely uncharted. Attest aims to address this specific, albeit narrow, problem by providing a mechanism to audit these external AI-generated descriptions.
Think of it less like a traditional compliance tool and more like a digital watchdog for your firm's public-facing AI persona. If your business relies on AI to generate marketing copy, sales pitches, or even answers to prospect queries, and there's no system to verify or record what that AI is saying, you're operating in a blind spot. Attest seeks to illuminate that blind spot.
How Attest Works: Auditing AI Outputs
Attest functions as an Agent Development Kit (ADK) agent. Its primary purpose is to query other AI models and then audit their responses concerning SEC-registered advisers. The process involves several key steps:
- Querying External Models: Attest is designed to interact with various large language models (LLMs) that might be used by or on behalf of an investment advisory firm. It poses questions or prompts that would typically be used in a client-facing scenario, such as asking about the firm's services, investment strategies, or compliance measures.
- Collecting Responses: All responses generated by these external AI models are captured and stored. This creates a historical record of what AI has communicated about the firm.
- Auditing for Accuracy and Fairness: The collected responses are then audited. This audit can involve several checks:
- Factual Accuracy: Does the AI's description align with the firm's actual services, disclosures, and regulatory filings?
- Fairness and Balance: Does the AI present a balanced view, or does it overstate capabilities or omit critical information? This is particularly relevant given the SEC's focus on fair claims.
- Consistency: Are the descriptions consistent across different queries and different AI models?
- Reporting: Attest generates reports detailing the AI's outputs, the audit findings, and any discrepancies or potential compliance issues. This provides a crucial audit trail for the firm.
The intention is not to replace the core function of these LLMs but to add a layer of oversight. For a financial institution, especially one registered with the SEC, trust and transparency are paramount. When AI becomes a significant channel for client interaction, ensuring that AI communicates accurately and ethically is no longer optional; it's a necessity. Attest provides a technical solution to a growing compliance and reputational risk.
Why This Matters: Regulatory Scrutiny and Future Implications
The SEC's focus on AI in financial services is a clear signal of the direction regulators are heading. While Attest is built to address a specific vulnerability – what external AI says about an adviser – it touches upon broader themes of AI governance and accountability in regulated industries. The SEC's 2026 priorities highlight a growing awareness of the potential risks associated with AI deployment. Firms that use AI to interact with clients must be prepared to demonstrate that these interactions are compliant, accurate, and do not mislead consumers.
The problem Attest tackles is the