Good Guys vs Bad Guys: AI-Powered Carbon Compliance Analysis

Research Overview

Corporate sustainability reports contain varying levels of compliance with carbon emission regulations. Our research uses artificial intelligence to analyze these reports and identify patterns of compliance and evasion across different company types.

The Challenge

While carbon pricing mechanisms and emission regulations expand across the European Union, corporate responses vary significantly. Some companies report transparently, while others employ tactics such as:

• Selective reporting of favorable data
• Regulatory arbitrage exploiting framework ambiguities
• Superficial initiatives without substantive change
• Creative accounting in emissions classification

Manual analysis of sustainability reports is time-intensive and difficult to scale, creating gaps in regulatory oversight.

Research Methodology

We’re using ChatReport, a tool for analyzing corporate sustainability reports against regulatory standards. Building on the ChatReport research methodology, we developed enhanced functionality for corporate compliance analysis.

Our ChatReport Implementation:

• Enhanced Question Sets: Customizable question frameworks for different analysis needs
• Web Application: Streamlit-based interface for uploading and analyzing sustainability reports
• Comparative Analysis: Tools for evaluating multiple companies against compliance criteria
• Data Visualization: Radar charts and CSV export for analysis results

The tool allows researchers to upload corporate sustainability reports and run systematic analysis using predefined question sets, generating comparable scores across different companies and compliance dimensions.

Technical Implementation: Our code and enhancements are available on GitHub at climateandtech/chatreport, including the web interface and customizable question set functionality.

Background Research: ChatReport is based on extensive academic research into LLM-based sustainability report analysis. For detailed information about the underlying methodology, see our comprehensive overview of the ChatReport study.

Sample Design:

• “Good Guys”: 15 companies recognized among the world’s most sustainable (Global 100 rankings)
• “Bad Guys”: 15 companies previously fined for carbon regulation violations

Analysis Approach: We built custom question sets on top of the ChatReport framework to systematically analyze sustainability reports from both company groups. Our custom question sets focus specifically on:

• Carbon emission disclosure quality and completeness
• Regulatory compliance indicators
• Transparency in reporting methodologies
• Verification and audit practices

This approach allows us to compare how different companies approach sustainability reporting and identify patterns that distinguish compliant from non-compliant behavior.

Preliminary Findings

Initial analysis shows measurable differences between company groups:

“Good Guys” tend to demonstrate:

• Robust governance frameworks integrating climate risks
• Comprehensive emissions reporting with external verification
• Clear accountability mechanisms

“Bad Guys” often exhibit:

• Inconsistent performance across TCFD criteria
• Selective reporting obscuring details
• Lack of external verification

Research Questions:

• RQ1: Statistical differences between compliant vs. non-compliant companies? Answer: Yes
• RQ2: What clusters exist beyond “good/bad”? Answer: Three clusters including “mixed” category
• RQ3: AI assessment confidence levels? In progress
• RQ4: Better evaluation criteria development? In development

Partnership with Viadrina University

We collaborate with Viadrina University (European University) research team, led by researcher Lolie Cambolive. This partnership provides expertise in European regulatory frameworks, corporate governance research, and machine learning applications.

Policy Applications

Our research supports evidence-based policy making by:

Identifying Regulatory Gaps:

• Loopholes enabling selective reporting
• Verification weaknesses in oversight systems
• Sector-specific vulnerabilities

Enhancing Enforcement:

• AI-powered monitoring tools for compliance assessment
• Automated flagging of inconsistent reporting practices

Open Science Approach

Research operates on open science principles with open datasets, transparent methodology, and collaborative platform for model improvement.

Next Steps

The integration of AI into regulatory oversight can help close gaps between corporate climate commitments and actual compliance. Our goal is developing practical tools for corporate climate accountability that provide reliable compliance assessments.

Join us in developing evidence-based tools for improving corporate climate accountability.