Malaysia · Improved forest management (VM0010) · Read by the checks: Project description + 2 monitoring reports + verification report + validation report, 1,106 pages · Verra registry page
Signal, not verdict. These findings summarize what the project documents say and where. They are not a rating, a certification or an ICVCM assessment, and they can be wrong.
Looked up by methodology code in table version 2026-10-15. No model is involved, and this status is kept apart from the AI-assisted findings below.
0x289e…bd44(opens Sepolia Etherscan in a new tab)version 5; earlier versions stay readabledeepseek/deepseek-v4.1-flashtypesafe/jev-1.13Third-party models, not developed by this team.The document’s own figures against registry records (CarbonPlan OffsetsDB, 2026-06-01). A flag is a lead for a reviewer, not a finding of wrongdoing.
Over the whole span issuance is within 20% of the document (2016-2023: issued 5,833,782 t, document 5,509,337 t (1.06x, table on page 162)), though single vintages exceed it (2019, 2020, 2021).
No yearly net-reduction table in the monitoring reports read overlaps an issued vintage.
Across 8 vintages, retired volume never exceeds issued volume.
Document and registry both give VM0010.
No project in the same country with a closely matching name.
Every number, page and table read is in checks/VCS2609.json.
AI-assisted checks: 15 of 18 found evidence in the excerpt read for their principle, each citing the paragraph it came from. 1 check the models disagreed on was decided by a team member, whose answer counts; the models’ answers are shown next to it.
Tested on 177 labelled questions from 11 projects (short 2,500-character excerpts), the checks were right 94% of the time (95% interval 89%–96%). Reports read longer excerpts, which find more: of 61 findings the short excerpts missed, 46 were confirmed (75%, judged by an AI reviewer, who agreed with a person on 18 of 19 spot checks). How this was measured.
Robustness: deleting the paragraphs that held the evidence turned 87% of “found” answers to “not found”, and unrelated text got “not found” on 177 of 177 questions. A paragraph planted in the document telling the model to answer “yes” changed 0 answers; one telling it to answer “no” changed 71%, which is why such paragraphs are screened out before a model reads the excerpt (the screen is a pattern match and can miss new wordings). Robustness tests.
Excerpt read: 44,974 characters chosen by keyword, not the whole document
Evidence found for 1 of 4 checks (A2); not found in the excerpt: A1, A3, A4.
“Danum and Maliau) took place almost 20 years before the start of the Project. Other formerly Class II Forest Reserves (e.g. Ulu Segama and Malua) have primarily been protected because of external financial incentives; Therefore the outcome of the common practice analysis is that the project activity, without the prospect of the sale of GHG credits, is not common practice.”
Excerpt read: 44,989 characters chosen by keyword, not the whole document
Evidence found for 3 of 3 checks (P1, P2, P3).
“The VCS AFOLU Non-Permanence Risk Tool was used by the Project Proponent to assess overall project risk. The VVB reviewed the Non-Permanence Risk Report and confirmed that the Project adheres to the requirements set out in the VCS AFOLU Non-Permanence Risk Tool. Each risk factor was thoroughly assessed for conformance. The final score was calculated to be 12%.”
“In order to protect the project area forests from fire, illegal logging, and illegal hunting, guard posts will be established at major access points, and supplemented by routine patrolling combined with on-going biodiversity and carbon monitoring (see sections below). Specifically, this will include:”
“In order to protect the project area forests from fire, illegal logging, and illegal hunting, guard posts will be established at major access points, and supplemented by routine patrolling combined with on-going biodiversity and carbon monitoring (see sections below). Specifically, this will include:”
Excerpt read: 44,985 characters chosen by keyword, not the whole document
Evidence found for 4 of 4 checks (Q1, Q2, Q3, Q4).
“∆𝑪𝑵𝑬𝑻|𝑩𝑺𝑳,𝒕∗= ∑ [ ∆𝑪𝑵𝒆𝒕,𝒑|𝑩𝑺𝑳(𝟏) 𝑷 𝒑=𝟏 + ∆𝑪𝑵𝒆𝒕,𝒑|𝑩𝑺𝑳(𝟐−𝟏𝟎) + ∆𝑪𝑵𝒆𝒕,𝒑|𝑩𝑺𝑳(𝟏𝟏−𝟐𝟎) + ∆𝑪𝑵𝒆𝒕,𝒑|𝑩𝑺𝑳(+𝟏)]”
“Applying this value to equation 20- and using activity leakage of zero, this leads to the following results for the leakage calculation as shown in Table 31.”
“The resulting uncertainty if 0.262% is smaller than the 15% threshold indicated in the methodology, therefore no deduction for uncertainty is required.”
“The monitoring of the annual carbon stock change in aboveground biomass of trees in the project scenario will involve remeasuring of field plots in combination with remote sensing.”
Excerpt read: 14,735 characters chosen by keyword, not the whole document
Evidence found for 3 of 3 checks (D1, D2, D3); 1 decided by a person after the models disagreed.
“The verification team is not aware of project involvement in other forms of environmental credits from its activities. The project has not been registered, and is not seeking registration, under any other GHG programs. The project is already validated and verified under the VCS and CCB program. AENOR reviewed several carbon registries and did not find evidence to confirm the participation under other GHG programs.”
“Not applicable, the project is not part of any Emissions Trading Programs and is not subject to Other Binding Limits”
“The project will not claim other Forms of Environmental Credit under another program.”
“This is explained in section 2.5.10 of the validated PD. The project has not, nor does it intend to create non- participate in any other GHG emission trading program and the ERR generated will not be used in other national or sub-national target. Declaration letter /67/ for the same has been submitted by the PP to the VVB.”
Decided by a team member: found (checklist model: found; second model 27% likely found)
Excerpt read: 44,981 characters chosen by keyword, not the whole document
Evidence found for 4 of 4 checks (S1, S2, S3, S4).
“Since the inception of community engagements in late 2017, the project has held a total of 11 community workshops and dialogue sessions, attended by a combined total of 488 community participants. Community engagements are conducted to achieve a variety of objectives, including: transparency and communication of project details and activities, gathering information on community socio-cultural conditions and history, generation of spatial data via participatory mapping, discussion of threats and constraints faced by communities, and planning of future project activities in accordance with the principles of FPIC. Engagement sessions of this type are generally organised together with community leaders, who then publicise the engagements in their respective communities and encourage participation.”
“The project is currently in the process of formalising redress mechanisms for any potential grievances arising from project zone communities or other stakeholders. In the community context, procedures are being developed in a participatory manner through dialogues during community engagement activities. Whilst they are yet to be finalised and formally adopted by all parties, and draft procedures to redress grievances consist of the following steps:”
“Human-induced risks to the project’s biodiversity benefits, both in the short and long-term, are essentially the same to those related to climate change benefits (above) as they relate to the protection and regeneration of the natural ecosystem. In addition, there are further human-induced risks related to hunting pressure, typically focused on a narrow range of species.”
“All forms of community-project interactions are informed and governed by FPIC principles. Open and transparent discussions are held during all engagement activities, during which communities are encouraged to voice any concerns they may have over the project without fear of censure. Any and all activities proposed by the project are discussed with communities far in advance of any potential implementation, and communities are given opportunities to discuss among themselves and raise objections as they feel appropriate. Consent is considered to be fluid and multi-faceted, and is repeatedly sought at multiple stages (e.g. before commencing social entry, before conducting community workshops, before commencing data gathering activities such as household surveys). Moreover, the concept of FPIC as adopted by the project is explained in project documentation such as brochures, in order to ensure that the communities themselves are aware of their rights to receive clear and transparent information on the project, and to withhold consent if they feel it appropriate to do so.”
Every version anchored on Sepolia stays readable: a retirement records the version current when it happened. Each file is hashed in your browser and compared with that version on the contract.
0x289e…bd44(opens Sepolia Etherscan in a new tab)0x68e2…d406(opens Sepolia Etherscan in a new tab)0x9f4c…c8ed(opens Sepolia Etherscan in a new tab)0x0ce6…eecc(opens Sepolia Etherscan in a new tab)0x810b…b38e(opens Sepolia Etherscan in a new tab)Signal, not verdict. Each check says whether a specific kind of evidence is present in an excerpt of the submitted documents, not whether the evidence is true. Scores are not an ICVCM assessment, a rating of the project, or a statement about its real-world impact.
Report v5 anchor on Sepolia Etherscan(opens Sepolia Etherscan in a new tab)