Executive Summary
MiFID II’s unbundling rules, introduced in 2018, fundamentally reshaped the European investment research market. The separation of research payments from execution commissions forced asset managers to assign explicit value to research for the first time, and triggered a significant restructuring of sell-side research production and distribution.
A second wave of research governance is now building — driven not by payment reform but by artificial intelligence. The questions that MiFID II raised about the value of research are being revisited in a new context: not just what research is worth, but what research is for, who is accountable for the conclusions drawn from it, and how those conclusions can be traced, verified, and defended.
This paper is a forward-looking analysis for investment professionals and research heads who want to understand where the regulatory arc is pointing — and who want to build research processes that will be ahead of it, not catching up to it.
1. The MiFID II Legacy: What It Changed and What It Left Unresolved
MiFID II’s research provisions required buy-side firms to pay for research explicitly — either from their own P&L or from research payment accounts funded by clients. The intent was to improve transparency in research valuations and reduce the conflicts of interest embedded in bundled commission arrangements.
The practical effects were significant: independent research firms found new commercial pathways; sell-side headcount contracted; and buy-side research budgets became explicit line items subject to management scrutiny. A decade of informal research economics was formalised.
What MiFID II did not address was the quality and traceability of research use. The rules defined how research should be paid for. They did not define what constitutes adequate research governance — how research should be used, documented, attributed, and preserved as part of the investment decision record.
This gap is where the next regulatory wave is forming.
2. The AI Catalyst: Why Governance Is Back on the Agenda
The proliferation of AI in investment research has reactivated the governance question in a more fundamental way than MiFID II’s payment reforms. When research consumption was human — an analyst reading a note, forming a view, writing a memo — the governance chain was short and largely implicit. When research consumption is AI-mediated — a platform retrieving, synthesising, and presenting conclusions from dozens of sources at once — the governance requirements change in kind, not just in degree.
Attribution at Scale
A human analyst citing a broker note creates a traceable evidence chain. An AI tool synthesising across twenty broker notes and presenting a consensus view may or may not preserve the provenance of each claim. If it does not, the compliance team cannot establish what the investment decision was based on.
Accountability Under Pressure
Investment decisions made on the basis of AI-generated synthesis need to be defensible. If a position goes wrong, the question ‘what was this based on?’ needs an answer. ‘The AI summarised the research space and suggested the thesis’ is not an answer that satisfies an Investment Committee, a regulator, or a client.
Entitlement and Rights
MiFID II created clarity on payment. It did not create clarity on use. As AI tools retrieve and process licensed content at scale, the question of what constitutes permitted use of research under bilateral agreements has become both more pressing and more contested.
3. The Regulatory Arc: Where the Rules Are Going
EU AI Act
The European Commission’s AI Act represents the most comprehensive regulatory framework specifically targeting AI systems. The majority of its substantive provisions — including rules on high-risk systems and Article 50 transparency requirements — enter application in August 2026. For investment firms operating in European markets, systems that use AI to inform investment decisions may fall within the high-risk category, triggering requirements around transparency, explainability, human oversight, and technical documentation.
FCA Approach
The UK Financial Conduct Authority has been explicit that it does not plan to introduce AI-specific regulations, preferring a principles-based, outcomes-focused approach that relies on existing frameworks. This means existing obligations around suitability, governance, conflicts of interest, and senior manager accountability apply to AI-assisted processes.
Sell-Side Engagement
Several major sell-side institutions have issued or are developing formal policies on AI use of distributed research. The direction is consistently toward requiring governance controls — specifically, no model training on distributed content, entitlement-aware retrieval, and audit logging — as conditions of AI-permitted use.
The regulatory arc is not pointing at AI in general. It is pointing at unaccountable AI — systems that make or inform decisions without leaving a traceable, reviewable evidence trail.
4. What Investment Research Governance Actually Requires
Explainability
The ability to explain, after the fact, what an AI-assisted research output was based on — which sources, which evidence, which reasoning steps. This is not the same as being able to describe how the AI model works in general terms. It is the ability to say, for a specific output: here is what the system retrieved, here is what it concluded, and here is why.
Human Oversight
AI research tools that produce outputs that flow directly into investment decisions without meaningful human review are exposed under most regulatory frameworks. The requirement is not that humans do all the work. It is that humans retain meaningful accountability for the conclusions.
Audit Trails
Durable, structured records of AI-assisted research activities. What queries were run, by whom, against what evidence universe, producing what outputs. These records need to be maintained for the same periods and under the same standards as other investment decision records.
Entitlement and Rights Compliance
AI research environments that retrieve licensed content must be able to demonstrate that retrieval is entitlement-aware and that content is not being used in ways that breach the bilateral agreements under which it is held.
5. Getting Ahead of the Curve
The investment teams best positioned as research governance requirements tighten are those that have established structured evidence governance as a standard operating practice — not as a compliance response to specific rules, but as a quality discipline embedded in the research process.
- Licensed-only retrieval with entitlement awareness
- Claim-level provenance preserved in every research output
- Structured audit logging of AI-assisted research activities
- Clear human review and accountability at the point where AI outputs inform decisions
Conclusion
MiFID II forced the buy-side to assign explicit value to research. The next wave of research governance will require firms to demonstrate the quality, provenance, and defensibility of the research they use — particularly where AI mediates that use.
The direction of regulatory travel is clear. The firms that get ahead of it will not be those that wait for precise rules. They will be those that establish structured evidence governance as a research process discipline — and build AI research infrastructure that is accountable by design.
Contours is a product of KiteEdge Ltd. To understand how Contours supports research governance in regulated investment environments, contact us.
