Research controls · AI-deployment disclosure

How AI-Assisted Research Is Governed

AegisIQ uses AI models to assist in producing equity research. This page names the controls that apply to that output before it reaches a subscriber.

1. Named human approval authority

The AegisIQ Research Review Officer holds final authority to approve, reject, or require revision of AI-assisted research before publication to subscribers. This authority belongs to the role, not merely to an individual; the role is currently held by H. Wayne Hayes Jr., Chief Executive Officer. A ticker request may retrieve an already approved, canonical research artifact. If no current approved artifact exists, the system may generate a draft, but that draft enters the review queue and is not published, exported, shared, or displayed as approved research until the Research Review Officer approves it.

2. Verification independent of the generating model

Published grades and valuation scenarios are computed by a deterministic, versioned scoring framework from identified data sources — a pipeline structurally independent of the AI language model that drafts narrative. Narrative is checked against that scored evidence and against vocabulary controls in review; the model does not verify itself.

3. Generation and approval records

Each generation and approval decision is recorded with the reviewer, role, decision, timestamp, methodology version, model version, source-manifest hash, artifact hash, and any revision note. Records are retained for at least seven years from creation, and longer when subject to a legal, regulatory, dispute, or preservation hold. Corrections are issued as new superseding records that reference the original; original records are preserved, not overwritten or deleted.

4. Publication boundary

Research follows a general-publication model: canonical by symbol, as-of date, and methodology version, with the same approved artifact available to all similarly entitled subscribers. Selecting a ticker selects a publication topic and does not personalize the analysis. Research does not use a subscriber's holdings, age, income, tax status, investment objectives, risk tolerance, projected cash flow, or personal circumstances. AegisIQ provides no portfolio-specific recommendation, allocation, suitability determination, individualized alert, trade instruction, discretionary authority, or automatic execution.

Full detail, including rating components, weights, grade boundaries, and correction procedure, is on the Methodology page. AI output remains subject to the limitations in the Investment Research Disclaimer.