VCM.fyi Forensic Intelligence Monitor

Forensic Intelligence

AI-powered weekly scanning of every carbon project across 100+ sources, delivering categorized intelligence memos with verifiable citations.

All Issued

Projects Scanned Weekly

100+

Sources Monitored

6

Intelligence Categories

Weekly

Scanning Cadence

The Forensic Intelligence Monitor is VCM.fyi's core capability. Every week, our AI engine scans every issued carbon project in our database across more than 100 sources — news outlets, regulatory filings, registry updates, scientific publications, and industry databases — to generate structured intelligence memos for each project.

See It in Action

The Problem

Carbon market professionals currently rely on periodic human assessments from rating agencies that update quarterly or annually. By the time a controversy, regulatory change, or market shift affects a project, traditional rating updates are weeks or months behind. A single missed event — a community conflict, a methodology dispute, a regulatory moratorium — can turn a "safe" credit into a reputational liability overnight.

How It Works

1

Multi-Prompt Architecture

Each project is scanned through four specialized prompts, each targeting a different intelligence domain: Article 6 bilateral agreements and ITMOs, CORSIA aviation offsetting status, carbon pricing mechanisms (ETS and carbon taxes), and VCM regulatory developments (export restrictions, moratoriums, offtake agreements). This multi-prompt approach ensures no intelligence category is missed due to prompt length constraints.

2

AI with Real-Time Web Search

Each prompt is executed using our proprietary AI engine with live web search grounding — the AI doesn't rely on training data alone. It actively searches the web during analysis, grounding every claim in current, verifiable sources. This means findings are timestamped and cited, not hallucinated from stale training data.

3

Semantic Deduplication

After scanning, a second-pass AI model performs semantic deduplication — identifying and merging findings that describe the same event in different words across different sources. This eliminates noise while preserving distinct findings that may appear similar but carry different implications.

4

Categorization and Delivery

Each finding is categorized into one of six intelligence categories: Integrity, Market & Offtake, Operational, Price Discovery, Regulatory, and Social & Community. Findings include source citations, market impact assessments, and "new this week" flags for fresh developments.

Data Sources

  • Real-time web search (live Google Search grounding)
  • Carbon registry updates (Verra, Gold Standard, ACR, CAR, Isometric, Puro Earth)
  • News outlets (Carbon Pulse, Reuters, Bloomberg, Environmental Finance)
  • Regulatory filings and government publications
  • Scientific publications and academic research
  • Industry databases and market reports
  • NGO and civil society reports
  • Project Design Documents (coming soon — PDDs will be scanned as an additional source for deeper project-level intelligence)

Technical Details

AI Engine

Proprietary AI with live Google Search grounding for real-time web access. Semantic deduplication uses a separate model pass optimized for cross-document similarity detection.

Processing Pipeline

Every Friday, all issued projects are scanned in parallel, deduplicated, and delivered to the platform — typically available within hours of scan completion.

Deal Flow Detection

The scanning prompts include specific markers for commercial activity: offtake agreements, forward purchase agreements, Letters of Intent (LOIs), Emission Reduction Purchase Agreements (ERPAs), and volume commitments. These are flagged under the Market & Offtake category.

Ground Truth Baselines

Each prompt embeds verified reference data to reduce hallucination: ICAO CORSIA 6th Edition (130 participating states), Article 6 Implementation Status Report (A6ISR) FY2024, and World Bank Carbon Pricing Dashboard. The AI cross-checks its search findings against these baselines.

Quality & Accuracy

  • Every finding includes a source citation with URL — users can verify any claim directly
  • Semantic deduplication removes 15-30% of raw findings that are duplicates phrased differently across sources
  • Ground truth baselines prevent the AI from contradicting verified reference data (ICAO CORSIA states list, World Bank carbon pricing data)
  • Weekly cadence means stale findings are naturally displaced by fresh intelligence
  • Market impact assessment provides a qualitative severity indicator for each finding

Update Frequency

Every Friday. Full scan of all issued projects with results available the same day. Historical findings are preserved — the platform shows the complete intelligence timeline for each project.

Explore forensic intelligence for any project

Start with our free tier — no credit card required.