The third use case in the 2022 Data Commons whitepaper: banks being pushed by regulation to report Environmental, Social, and Governance (ESG) data they largely cannot get from their own customers.
The regulatory push
The European Banking Authority (EBA), to support the European Commission’s action plan on financing sustainable growth, is building ESG data into the regulatory/supervisory framework for EU credit institutions — mandatory, consistent disclosure requirements including granular templates, tables, and instructions, aimed at fixing inconsistent current ESG disclosures.
The data gap this creates
Banks must report ESG-relevant information in their disclosures, but the proposed framework doesn’t require all of banks’ customers to report the underlying data. Large corporations disclose some of it publicly; data from SMEs largely stays unknown — some SMEs simply don’t have the information, others lack the IT capacity/resources/knowledge to gather it. Even where data could be shared, SMEs are wary of exposing sensitive information.
This matters beyond disclosure optics: with climate stress-tests underway, managing climate risk carries indirect (and eventually direct) capital implications for banks, feeding into credit-pricing and allocation decisions.
Why the whitepaper frames this as a Data Commons case
In a setting with a wide range of stakeholders with different interests (banks, large corporates, SMEs, regulators), the paper argues a Data Commons approach has an advantage over proprietary solutions — the same Reciprocity/Security/Openness logic that’s meant to get reluctant, resource-constrained SMEs to participate without exposing sensitive data.
The greenwashing feedback loop (Upright, 2025)
Added 2026-09-10. A finding worth carrying on this page — with its weight stated: it rests on a single self-published, non-peer-reviewed study run on the author’s own closed benchmark, with no independent replication (evidentiary health warning). Evaluating eleven frontier LLMs on impact-magnitude judgements, Upright reports that they “often favor contrafactuals that cast companies’ impacts in a more positive light and minimize especially environmental impacts”, and attributes this to training corpora containing “large volumes of corporate sustainability reports and similar content” (Ojala, 2025). Stated as a suspicion, not a demonstration.
⟨Inference: that closes a loop. Corporate self-reporting enters the training corpus; the corpus trains the models; the models are then sold to assess corporate sustainability; their output becomes further reporting. Greenwashing entering the epistemic base and coming back out laundered as model knowledge. Upright names the bias but not the loop.
For this page it is the strongest available argument that attested primary data and inferred data are not interchangeable, however good the inference gets — an ESG data market built on models trained on self-reports has a systematic optimistic bias no amount of model quality removes. See Verifiable carbon accounting and Decision-grade carbon-data criteria.⟩
Tension held open ⟨lint 2026-09-14⟩. As written, the paragraph above is absolutist, and the wiki’s own standing position is not. Criterion H of Decision-grade carbon-data criteria holds that a decision-grade system “does not require perfect primary data for every input on day one” but a rule-governed pathway that distinguishes measured, supplier-specific, modelled and default data, discloses the quality of each, and improves over time; PACT’s data-quality regime likewise scores secondary and estimated data rather than excluding it. Those two claims are only reconcilable if “inferred” here means LLM judgement inferred from a corpus of self-reports and not modelled data with a disclosed quality rating — a distinction the wiki does not currently draw anywhere. Until a source settles it, both readings stand: the narrow one is defensible and the broad one would put this page against three others.
SFDR and CSRD assessed against circularity (EEA, 2026)
The EEA’s circular-finance report tests two of the instruments this page tracks against one specific subject matter, and both fail the test (§3.1.1):
- SFDR disclosure standards “do not fully allow the benefits of circular businesses to be quantified. Consequently, the benefits cannot be factored into financing decisions.”
- CSRD reporting on circular economy and resource use is “still immature”; it is also “hard to disaggregate circular businesses in bonds and funds and in banking data.”
Its remedy is the one this page already treats sceptically: the CSRD review “could reduce information asymmetry for investors by standardising the financial statements that companies are required to produce so that they are relevant and comparable” (§4.2) — a disclosure answer to a measurement problem, with no statement of where the underlying data comes from or how it is checked.
⟨Inference: this is the same disclosure-ahead-of-the-data-layer shape the wiki records for carbon and for biodiversity. Third domain, same order of operations: the reporting obligation is written first, the primary record it would draw on is assumed.⟩
Cross-references
- Data Commons
- Deutsche Bank — co-author, brings the banking-sector angle
- Evaluating LLMs (Ojala, 2025) — the greenwashing-in-corpus finding
- Upright Project — a commercial impact-data vendor in this market
- Unlocking the circular economy (EEA, 2026) — SFDR and CSRD measured against circularity
- Circularity definition problem · Circularity valuation gap
Sources
- sine-data-commons-2022 —
sources/whitepapers/2022-04-25-data-commons-final.pdf sources/circular-finance/2026-11-05-eea-unlocking-the-circular-economy/2026-11-05-eea-unlocking-the-circular-economy.md, §3.1.1, §4.2. EEA Report 02/2026, doi:10.2800/5422329.