Investment Banking AI: MiFID II Research Unbundling, SEC Research Analyst Rules & AI Governance
Investment banks deploying AI across M&A advisory, equity research, trading, and capital markets operations face a complex regulatory landscape spanning MiFID II's research unbundling requirements, SEC research analyst rules under Regulation AC, FINRA supervisory obligations for algorithmic trading, and information barrier requirements that AI systems can inadvertently compromise. The AI governance stakes in investment banking are amplified by the potential for market-moving information flows across AI systems that have no concept of Chinese walls.
MiFID II Article 13 — Research Unbundling and AI-Generated Research
Regulation: Markets in Financial Instruments Directive II (MiFID II), Directive 2014/65/EU; implemented January 2018
Key provision: Article 13 of MiFID II Delegated Directive prohibits investment firms from receiving research as an inducement bundled with execution services — research must be separately priced and paid for
AI question: When AI systems generate market commentary, trade ideas, or sector analysis for institutional clients, regulators are examining whether such output constitutes "investment research" subject to unbundling rules
ESMA guidance: ESMA has indicated that AI-generated content meeting the definition of investment research under MiFID II is subject to the same rules as human-produced research
Enforcement: FCA has issued findings on research payment account compliance; non-compliant bundling creates inducement liability under MiFID II Article 24
SEC Regulation Analyst Certification (Reg AC) — AI Research Certification Problem
Rule: Regulation AC, 17 C.F.R. § 242.500 et seq.
Requirement: Research analysts must certify that their reports accurately reflect their personal views and disclose whether compensation was tied to specific recommendations
AI compliance gap: AI-generated research reports have no "analyst" to certify — creating a gap in the Reg AC certification framework that the SEC has not yet addressed with formal guidance
Current approach: Banks are generally treating AI-generated content as "market commentary" rather than "research" where possible to avoid Reg AC requirements — but regulators are examining whether this characterization is accurate for substantive AI investment analysis
Risk: Mischaracterizing AI research as commentary to avoid Reg AC creates separate disclosure liability
AI and Information Barrier Compliance
Investment banks maintain information barriers (Chinese walls) between their advisory/M&A divisions and their trading/research divisions to prevent material non-public information (MNPI) from flowing from advisory contexts into trading decisions. AI systems that are trained on, or have access to, data from across the enterprise create an information barrier compliance risk that legacy Chinese wall controls were not designed to address.
Claire's Investment Banking AI Compliance Solution
Claire for Investment Banking AI Governance
AI Research Classification
Claire's content classification module analyzes AI-generated market content and applies MiFID II and Reg AC classification logic — distinguishing between investment research (subject to unbundling and certification requirements), investment recommendations (subject to suitability obligations), and general market commentary. Classification outputs feed into compliance workflows before client distribution.
Information Barrier Monitoring for AI Systems
Claire audits AI model training datasets and access permissions to identify potential information barrier violations — including cases where AI systems used in public-side trading operations have access to data from private-side advisory engagements. Data lineage tracking ensures Chinese wall integrity is maintained at the AI system level.
Algorithmic Trading Supervision
Claire provides FINRA Rule 3110 supervisory framework support for AI-driven trading algorithms, including pre-deployment review documentation, ongoing monitoring of algorithmic trading patterns, and exception reporting for behaviors that may constitute market manipulation or violate best execution obligations.
Investment Banking AI Compliance Checklist
MiFID II / SEC / FINRA AI Compliance for Investment Banks
AI research classification policy: Written policy governing when AI-generated content constitutes investment research under MiFID II and Reg AC, with classification applied before client distribution.
Information barrier review of AI training data: AI systems used in public-side operations audited to confirm they have not been trained on, or have access to, MNPI from private-side advisory engagements.
MiFID II inducement analysis for AI tools: AI tools provided to institutional clients reviewed for potential inducement classification — AI tools that drive order flow to the providing firm may constitute bundled research.
Algorithmic trading pre-deployment review: All algorithmic trading systems reviewed against FINRA Rule 3110 supervisory requirements and SEC market manipulation rules before production deployment.
Best execution AI monitoring: AI-driven order routing systems monitored against MiFID II best execution requirements (Article 27) with documented evidence that execution quality is prioritized.
AI model inventory for trading systems: Complete inventory of all AI systems involved in trading decisions, with SR 11-7 model risk management governance applied to material models.
Reg AC compliance framework for AI-assisted research: Written framework governing how Reg AC certifications apply to research that is AI-assisted vs. AI-generated, reviewed with outside counsel.
Market manipulation monitoring: AI trading systems monitored for patterns that may constitute layering, spoofing, or other manipulative trading — subject to ongoing surveillance under FINRA Rule 3120.
Communication surveillance for AI-assisted messaging: AI-assisted client communications reviewed under existing electronic communication surveillance programs — AI-generated content does not exempt communications from recordkeeping and surveillance obligations.
Board-level AI governance: Board risk committee oversight of AI use in capital markets activities, with quarterly reporting on model performance, compliance findings, and emerging regulatory developments.
Frequently Asked Questions
Is AI-generated market analysis "investment research" under MiFID II?
ESMA has confirmed that the classification of content as investment research depends on its substance, not the method of production. AI-generated content that provides investment recommendations or analysis of financial instruments on a regular basis, distributed to multiple clients, meets the MiFID II definition of investment research and is subject to unbundling requirements. Banks cannot avoid MiFID II research rules by labeling AI-generated research as "market commentary."
How does Reg AC apply to AI-assisted equity research reports?
Reg AC requires research analysts to personally certify their reports. For AI-assisted reports where a human analyst reviews and endorses the AI-generated analysis, the certifying analyst must be able to truthfully certify that the report reflects their personal views. If the analyst substantially defers to AI analysis without genuine review, the certification may be false. Banks should require genuine analyst review before certification of AI-assisted reports.
Can an investment bank use AI for M&A deal sourcing?
Yes, but AI systems used for deal sourcing must be isolated from the bank's trading and research operations under information barrier policies. AI that identifies potential M&A targets using public data is generally permissible. AI that has access to confidential engagement data must be governed under the same Chinese wall policies as human M&A professionals, with access controls and audit trails.
What FINRA rules apply to AI-driven trading algorithms?
FINRA Rule 3110 requires supervision of all trading activities, including algorithmic trading. FINRA has issued multiple guidance documents on algorithmic trading supervision, including Regulatory Notice 15-09, which requires pre-deployment review of algorithms, ongoing monitoring, and emergency shutdown procedures. AI trading algorithms must be subject to these supervisory requirements before deployment.
How are MiFID II best execution requirements applied to AI order routing?
MiFID II Article 27 requires investment firms to take all sufficient steps to obtain the best possible result for clients in order execution, considering price, cost, speed, likelihood of execution, and other relevant factors. AI order routing systems must be calibrated to optimize these factors for client benefit — not for the firm's own execution economics. Firms must maintain execution quality data and provide annual best execution reports.
Related: Finance AI Overview | MiFID II AI Compliance | Trade Surveillance AI