This shift is not happening in isolation. Messaging from European regulators, including ESMA, the ESAs and the ECB, points in the same direction. Market stress risks remain elevated, vulnerabilities in private markets are increasing and the rapid adoption of AI is introducing new layers of complexity.
For boards, the implication is straightforward. Oversight expectations have moved beyond understanding outputs. Regulators now expect boards to challenge the assumptions, data and models that sit behind them.
What has shifted in 2026
The regulatory emphasis has evolved in a noticeable way.
Inflation and interest rate risk have eased sufficiently to fall back as standalone concerns. In their place, other risks have intensified and now demand greater board attention:
- Asset valuation risk, particularly in less transparent markets
- Ongoing market and systemic stress risk
- Data quality, modelling and the use of AI
At a European level, the message is consistent. Supervisors are increasingly focused on resilience under stress, rather than stability in benign conditions. There is also a clear concern around opacity, particularly in private credit, private equity and other hard-to-value assets.
This is not a technical recalibration. It is a shift in supervisory mindset. Boards should read it as part of a broader direction of travel rather than a standalone update.
Valuation: from pricing exercise to governance framework
The CBI’s concerns on valuation are both specific and practical.
It highlights stretched valuations across equity and debt markets, alongside the potential for rapid changes in sentiment. Particular attention is drawn to private assets, where transparency is limited and pricing often depends on models or expert judgement.
There is also a forward-looking dimension. The outlook notes that AI-related valuations could reprice sharply if expectations around adoption or profitability shift.
In response, the CBI has announced a 2026 thematic review of hard-to-value assets, including private equity, private credit and other illiquid securities. The focus will be on governance, policies, models and controls.
This builds on earlier supervisory findings, which identified recurring weaknesses:
- Valuation policies that are incomplete or poorly documented
- Unclear accountability for model validation
- Limited differentiation between normal and stressed market conditions
- Weak linkage between liquidity stress testing and valuation practices
Taken together, these findings point to a consistent issue. Where reliable market prices are absent, valuation becomes less about calculation and more about governance.
For boards, the key question is not whether valuations are produced, but whether they are controlled, evidenced and challengeable.
Data, modelling and AI: governance is the real challenge
Alongside valuation, the CBI places equal emphasis on data, modelling and AI.
The increasing use of third-party AI tools and advanced models is changing how decisions are made across financial institutions. While this brings efficiency and analytical power, it also introduces new risks, particularly where governance frameworks have not kept pace.
One of the more direct messages from the CBI is that poor data quality is rarely just a systems issue. More often, it reflects weaknesses in governance, ownership and resourcing. Short-term fixes and workarounds remain common, but supervisors are increasingly focused on longer-term structural improvements.
This concern is echoed at European level. ESMA continues to push firms on valuation practices and AI usage in investment services. The ECB has maintained risk data aggregation and reporting as a supervisory priority. Recent supervisory assessments show that progress in these areas remains uneven, with recurring issues including weak data architecture, insufficient controls and limited board-level prioritisation.
For boards, this shifts the conversation materially. It is no longer enough to ask whether a data or model programme exists.
The more relevant questions are:
- Can the board see the quality of the data underpinning key decisions?
- Which valuations or models rely on degraded or incomplete inputs?
- Is the pace of remediation sufficient to change the risk profile?
Without clear answers to these questions, oversight remains partial.
What this means for boards
An effective board response should be both integrated and proportionate.
It should be integrated because valuation risk, model risk and data risk are closely linked. They often depend on the same inputs, assumptions, third parties and reporting structures. Addressing them in isolation creates blind spots.
It should be proportionate because the nature of the risk differs across sectors. Banks, funds, SPVs and other financial institutions will each face different manifestations of the same underlying issues.
Despite these differences, supervisory expectations are consistent. Boards are expected to demonstrate:
- Clear accountability for valuation, model and data governance
- Evidence-based challenge, supported by meaningful information
- Transparent escalation of issues and risks
- Robust oversight of third-party providers and delegates
- Credible and well-documented remediation plans
The underlying message is simple. Governance frameworks must be capable of withstanding scrutiny, not just producing outputs.
Sector-specific focus areas
While the themes are common, their application varies by sector:
| Sector | Board focus | What can go wrong |
| Banks | Risk data aggregation, internal models, AI in decision-making, model-change governance. | Fragmented data, model drift, weak validation, poor explainability and manual workarounds. |
| Funds / SPVs | Hard-to-value assets, pricing, investor disclosure, transaction data, delegated oversight, reporting | Stale pricing, over-reliance on judgement, weak validation, pricing opacity, poor error formalisation, reliance on servicers, administrators, trustees and external data with weak end-to-end challenge. |
| MiFID / Payments & E-Money | Fraud and customer outcome models, reliance on third party/vendor solutions, data quality | Model opacity, underestimation of tail risks or scenario impacts, over-reliance on expert judgement. |
| Insurance | Pricing, underwriting, claims | Operational Models – Lack of model inventory, weak accountability, limited explainability, weak ethics/fairness considerations. Financial reporting models – Insufficient challenge to modelled results, lack of appreciation of uncertainty around future outcomes, out of date model governance. |