Digital MRV is moving from a technical detail to an integrity question. In September 2026, an expert working group co-managed by the Integrity Council for the Voluntary Carbon Market (ICVCM) and the World Bank released a report arguing that the satellite, sensor and AI tools used to quantify emission reductions should be approved before use, re-approved when their training data changes, and have their calculation procedures disclosed. For buyers of voluntary carbon credits, the signal is clear: the quality debate is shifting from individual projects to the software that measures them.
The Core Carbon Principles Stay, the Guidance Does Not
The report, a 46-page Version 1.0 document dated September 2026 and introduced on the ICVCM website on September 22, is the product of six meetings held between March and July. Its central conclusion is that the ICVCM’s ten Core Carbon Principles, the quality benchmark behind the CCP label, do not need to be rewritten for the digital era. What is needed, the group argues, is additional guidance to apply those principles consistently when measurement is done by machines rather than field teams.
The working group brought together crediting programs, digital MRV providers, development finance institutions, verification bodies and their accreditation organisations, indigenous peoples’ representatives and NGOs. Named participants include Verra, Gold Standard and Climate Action Reserve, along with forest carbon data firm CTrees and ratings agency Sylvera.
One framing in the report deserves attention from both sides of the technology debate. Digital MRV and conventional MRV, it states, are both models of reality with their own uncertainties, and neither is inherently superior. A field team measuring sample plots and extrapolating an entire forest is doing the same kind of estimation a satellite does. The relevant question is not which tool is used, but whether its error margins and accountability are understood.
What 137 Market Participants Are Worried About
The recommendations rest on a survey conducted between December 2025 and January 2026, which drew 137 responses from market participants. On the benefits side, around three quarters cited transparency and more than half pointed to improved accuracy as advantages of digitalisation.
The concerns are concentrated. Data quality topped the list at 58% of respondents, followed by bias in algorithms or models at 39% and the lack of visibility into the tools themselves at 38%. Those three numbers map directly onto the report’s 16 recommendations, which span definitions, cybersecurity and conflicts of interest, disclosure, verification and data quality, and the rights of indigenous peoples and local communities.
Approval, Re-Approval and Disclosure: The Operational Core
The heart of the report is a proposal to bring tools, which today sit outside the rules, into the scope of approval, disclosure and verification. Crediting programs would approve a digital tool before it is used in credit issuance, against published criteria covering technical accuracy, eligible geographies and project types, and an explicit statement of what the tool can and cannot do.
Approval would not be one-off. The report lists at least four triggers for re-approval: major changes to the tool, changes to training data or settings, use outside the approved scope, and the simple passage of time. Methodology approval would also change. If a methodology uses a tool to calculate credit volumes, that tool becomes part of the assessment, including its treatment of uncertainty and bias and, where possible, evidence checked against field measurements rather than the vendor’s internal testing alone.
On disclosure, programs would publish a tool’s calculation procedures, training data characteristics, inputs, assumptions and conditions of use. Where trade secrets or privacy law block full publication, the report demands a separate mechanism providing an equivalent level of assurance, itself documented and subject to third-party review. It also asks the ICVCM to build minimum templates that separate detailed disclosure for verifiers from a summary for the general public.
Verification bodies face their own upgrade. They would need to demonstrate the capacity to audit tools, including tracing data provenance and inspecting algorithms. Where a tool cannot quantify its own uncertainty, credits would be discounted conservatively. And programs would decide in advance what happens when a tool is manipulated or malfunctions: suspension of issuance, quarantine of affected credits while investigations run, and predefined treatment of credits already issued or used.
Why One Model Error Is Now a Systemic Risk
The rationale for all this is propagation. A single satellite model or AI system can serve many projects across different countries and crediting programs, so one error lands on many credits at once. Shared methodologies and default values have always carried that risk, the report notes, but those are public documents anyone can read and challenge. Proprietary tools are not, and the technology companies that build them currently sit outside the accountability framework.
Nothing Changes Yet: Three Markers to Watch
The report is explicit about its own limits. It is a summary of the working group’s opinions, not the position of the ICVCM, and it did not go through a formal consensus process. Credit assessment and issuance are unchanged as of today. The ICVCM says it will treat the report as input for reviewing its Assessment Framework and would run a separate consultation before any change.
Three markers from here. First, whether the ICVCM puts a draft revision of the Assessment Framework out for public comment. Second, whether programs such as Verra and Gold Standard begin publishing their tool approval criteria and management mechanisms on their own. Third, how the market defines the equivalent level of guarantee demanded for proprietary tools: that definition will decide whether disclosure becomes a real obligation or a comfortable loophole.
